either-float and the syntax of co-€¦ · ellipsis approach to displacement of either (cf....

MARCEL DEN DIKKEN

EITHER-FLOAT AND THE SYNTAXOF CO-OR-DINATIONw

ABSTRACT. The syntax of either . . . or . . . treats the analyst to two main puzzles:the ‘either too high’ puzzle (either can be dissociated from the contrastive focus,

surfacing in positions higher up the syntactic tree), and the ‘either too low’ puzzle(either is apparently too low in the tree, embedded inside the first disjunct). Coveringdata beyond the range of extant accounts, this paper presents an integrated solution

to both puzzles. The paper’s central claim is that both either and or are phrasalcategories. They originate in a position adjoined to their disjunct, to the contrastivefocus or to a higher node on the ‘h-path’ projected from the contrastive focus.

Though either itself is immobile, its [+NEG] counterpart, neither, can undergophrasal movement from its base-generation site to a higher position in the tree, fromwhich it triggers negative inversion; [+WH] whether must move to SpecCP; and orand [+NEG] nor must, if they are not base-generated there, front to the initial

position in the second disjunct in order to be able to participate in a feature-checkingrelationship with the abstract head J(unction), the functional head that takes thesecond disjunct as its complement and the first disjunct as its specifier. The move-

ment of whether, neither and (n)or will be diagnosed on the basis of the familiarrestrictions on movement.

1. INTRODUCTION

The syntax of either . . . or . . . treats the analyst to two main puzzles,to which many of the finer-grained peculiarities of this disjunctionconstruction can be reduced. The two puzzles are illustrated in (1)

w Part of the material for this paper was first presented in a seminar co-taughtwith Janet Fodor at the CUNY Graduate Center in the spring of 2003. An imme-diate predecessor of the present text was delivered at the 28th GLOW Colloquium in

Geneva, Switzerland, on 1 April 2005. I thank the audiences present on bothoccasions for their generous feedback. I am very grateful to Aniko Liptak forcarefully reading and meticulously commenting on earlier versions of the paper.

Many thanks are also due to the anonymous reviewers for NLLT, as well as to editorPeter Culicover. Their invaluable input has led to substantial improvements of theanalysis and to refinements of the empirical material. The responsibility for the final

product is mine alone.

Natural Language & Linguistic Theory (2006) � Springer 2006DOI 10.1007/s11049-005-2503-0

and (2). The problem in (1), instantiating the ‘either too high’ puzzle,is that either seems to be able to ‘run away’ from the focus of the firstdisjunct (here, the object rice ) and surface in positions higher up thesyntactic tree. That is, either in (1b,c) is ‘higher than it is supposed tobe’. The problem illustrated in (2) (the ‘either too low’ puzzle) is theconverse of that in (1): either in (2b) is apparently ‘too low’ in thetree, ‘buried’ inside the first disjunct, which, by the Law of Coordi-nation of Likes, would have to be a full-fledged IP to match thesecond disjunct. Appropriating Kayne’s (1975) terminology forQ-float (‘L-tous’ and ‘R-tous’), I will refer collectively to both phe-nomena as ‘either-float’, and, more specifically, to instances of eitheroccurring further to the left than might be expected as ‘L-either’ and tocases of either occurring further to the right as ‘R-either’ (not meaningto prejudice the analysis thereby; the terminology adopted is meant tobe pretheoretical).

(1)a. John ate either rice or beans.

b. John either ate rice or beans. ~L-either

c. Either John ate rice or beans. ~L-either

(2)a. Either John ate rice or he ate beans.

b. John either ate rice or he ate beans. ~R-either

These two puzzles are familiar from the literature, though substan-tially more ink has been spilled over the first than over the second (cf.esp. Larson l985; Schwarz 1999 for detailed discussions of L-either).For (1), both movement and ellipsis accounts have been proposed.For (2) it is plain that an ellipsis analysis is unformulable (seeHendriks 2003:33–34 for an explicit demonstration), while a move-ment-based approach that assumes (2a) as the underlying represen-tation (on the basis of the standard idea that the first disjunct cannotbe smaller than the second; but see Larson 1985 for an ‘unbalanced’coordination approach to (2b)) has to grapple with the unwieldydownward movement of either that then seems required to derive(2b).

Without denying that movement and ellipsis play a key role in thesyntax of either . . . or . . . , the present paper’s main innovation is itsargument to the effect that both either and or are phrasal categories.This entails that neither either nor or is itself a disjunction particle.The phrasality of either will be shown to lead, in conjunction with the

MARCEL DEN DIKKEN

hypothesis (pioneered in Hendriks 2001, 2003) that the surface dis-tribution of either is strongly tied to contrastive focus, to an inte-grated solution to the puzzles in (1) and (2). Section 3 of the paper isdedicated to showing this in detail. Either will be shown to beimmobile (cf. also Han and Romero 2004, contra Larson 1985); buteither’s negative and [+WH] incarnations, neither and whether, dohave the ability or the obligation to move, as section 4 will show, onthe basis of the familiar restrictions on movement. Either’s coun-terpart (n)or in the second disjunct is systematically forced to be atthe left edge of the second disjunct, where it ends up either via base-generation or (if it does not originate on the left edge) via move-ment: (n)or is not a disjunction particle but a phrasal element thatneeds to establish a local, feature-checking Agree relationship withthe abstract functional head J in the structure of coordinationconstructions in (3) (a structure familiar from the recent literature;cf. Munn 1993; Kayne 1994; Johannessen 1998, and references therein).

(3) <either> (. . .) [JP [XP (. . .) <either> . . .] [J [YP or . . .]]]

Section 5 puts the structure in (3) and the approach to either . . . or. . . that it reflects in a broader perspective, extending it to both . . .and . . . and considering the motivation for the left-peripheralplacement of or and and in the second dis/conjunct. Finally, sec-tion 6 offers a brief summary of the major claims and accom-plishments.

2. L-EITHER : A LOOK AT THE LITERATURE

The principal focus of this paper is the phrasality of either and or, notthe ‘either too high’ puzzle in (1), which is only one piece of the bigpuzzle presented by the syntax of either . . . or . . . disjunction con-structions. But to set the stage, it will be good to start out thediscussion with a quick look at the previous literature on what I amcalling L-either – if only because, with the notable exception ofHendriks (2001, 2003), the literature on the syntax of either. . . or. . .constructions has focused primarily on the puzzle in (1): the fact thateither can surface in a position that is ‘too high’, higher than theposition which it would seem to originate in (i.e., the position rightnext to the contrastive focus, as in (1 a)). For the paradigm in (1), two

EITHER-FLOAT AND THE SYNTAX OF CO-OR-DINATION

main types of approach are represented in the literature: a movementanalysis (defended in Larson 1985; Munn 1993; cf. (4a)) and anellipsis approach (championed by Schwarz 1999; cf. (4b)).

(4)a. <Eitheri> John <eitheri> ate <eitheri> [rice or beans].

b. John ate either [[NP rice] or [NP beans]].John either [[VP ate rice] or [VP ate beans]].Either [[IP John ate rice] or [IP John ate beans]].

In defence of his movement approach, Larson capitalises onthe island effects illustrated in (5)–(7), with (5) instantiating an‘inner island’ effect,1 (6) a CNPC effect, and (7) a wh-islandviolation.2

(5)a. (?)John didn’t eat either rice or beans.

b. ??John either didn’t eat rice or beans.

c. ??Either John didn’t eat rice or beans.

1 Note, though, that ‘inner island’ (or negative island) effects typically do not give

rise to marginality but instead to ungrammaticality: thus, How strongly don’t youthink inflation will rebound? blocks a ‘downstairs’ reading for how strongly categor-ically.

2 I should note that (7b) is not rejected outright by all native speakers I have asked.But to most, (7b) does indeed sound awkward. It seems to improve substantially,however, with to preceding retire: cf. (i), which many speakers accept readily.

(i) John was either wondering whether to resign or to retire.

This observation (which to my knowledge has not been reported in the literaturebefore) is something that no extant account of the either. . . or. . . construction seems

to have a handle on. Note that there is no general restriction that prevents VPs in thecomplement of to from being foci (cf. What John wanted to do was (to) retire). Butinterestingly, when the focused infinitival VP is inside a whether-CP, some speakersdo suddenly report a marked preference for to: both variants of (ii) are slightly

awkward due to the weak wh-island effect induced by extraction of what acrosswhether, but the variant without to is deemed worse by some of my informants. Anellipsis approach to displacement of either (cf. Schwarz’s 1999 analysis, discussed

below) could accommodate these facts with the aid of (iii) (Wilder 1997) once it isensured that, in whether-infinitives, the focus is the to-IP. It will remain to bedetermined, however, why it ostensibly cannot be VP.

(ii) ?What John was wondering whether to do was ??(to) retire.

(iii) Ellipsis goes all the way down to but never into the focus.

MARCEL DEN DIKKEN

(6)a. John revised [NP his decision [CP C [IP PRO to cook eitherrice or beans]]].

b. *John revised either [NP his decision [CP C [IP PRO to cookrice or beans]]].

(7)a. John was wondering [CP whether [IP PRO to either resignor retire]].

b. *John was either wondering [CP whether [IP PRO to resignor retire]].

The locality conditions on L-either are stricter than expected on anisland-based analysis, however – any CP blocks it, even CPs whichare not islands for garden-variety A¢-extraction.3

(8)a. <Either> John <either> wanted to eat <either>rice or beans.

b. <??Either> John <??either> wanted <%either> for youto eat <either> rice or beans.

c. <??Either> John <??either> said <%either> that hewould eat <either> rice or beans.

This has led Munn (1993) to propose an analysis of L-either in termsof S-structure Quantifier Raising (QR) – an analysis which bothcapitalises on the fact that either doubles as a quantifier (as in Eitheranalysis will capture the facts) and immediately assimilates the facts in(8) to the scope facts in (9).

(9)a. Some girl wanted to watch every 007 movie. X">$

b. Some girl wanted for you to watch every 007 movie. *">$

c. Some girl said that she would watch every 007 movie.*">$

Schwarz (1999) counters Munn’s (1993) ‘overt QR’ approach toL-either, however, by pointing out that L-either is blocked incontexts in which inverse scope is not. The key data, involvingverb–particle constructions, are reproduced here in (10) and (11),

3 While Larson (1985) and Schwarz (1999:342) report that sentences like (8c) arebad with either placed immediately to the left of the that-clause, there are speakerswho accept this; a similar observation holds for (8b). See section 3.5 below, for

discussion.


with the left-hand examples in (10) being unacceptable forSchwarz’s informants (though Han and Romero 2004:fn. 12 note,attributing the observation to a reviewer, that there is speakervariation on this point, ‘with emphasis, in certain discoursecontexts’).

(10)a. ??Either this pissed Bill or Sue off. vs.Either this pissed Bill off or Sue.

b. ??Either they locked you or me up. vs.Either they locked you up or me.

c. ??Either he gulped one or two down. vs.Either he gulped one down or two.

(11)a. Something pissed every guest off.

b. Some sheriff locked every gangster up.

c. Some boy gulped every coke down.

The crucial thing to note here is that the examples in (11 a–c) all allowan inverse scope reading (">$) (Schwarz 1999:349; their counter-parts with the particle adjacent to the verb support this reading rathermore readily, though). This then compromises the QR approach toL-either – after all, in exactly the same context in which L-either isimpossible in (10), QR is nonetheless possible.4

Schwarz (1999) argues that an ellipsis account does make sense ofthe deviance of the left-hand examples in (10). He captures this with acombination of an appeal to the general requirement that gappingnot leave ‘dangling remnants’ (cf. (12)), which ensures that therepresentation in (13a) is ungrammatical, and an appeal to RightNode Raising, which, when taking (13b) as its input, delivers a

4 One might seek to salvage the movement analysis, however, by assimilating thefacts in (10) to cases of subextraction from the sandwiched noun phrase; cf. Who did

he look <up> the number of <??up>?

MARCEL DEN DIKKEN

suboptimal output: RNR of a bare particle up to a position above IPis highly marginal (cf. Schwarz 1999:359).5

(12)a. Some talked *(with you) about politics and others talkedwith me about music.

b. John dropped the coffee and Mary (*clumsily) droppedthe tea.

(13)a. *Either [IP this pissed Bill] or [IP this pissed Sue off].

b. Either [IP this pissed Bill off] or [IP this pissed Sue off].

Schwarz (1999:359–360) supports his RNR-based approach to theleft-hand examples in (10) by pointing out that their non-ellipticalcounterparts are likewise degraded with either in sentence-initialposition, as seen in (14). He also suggests that the grammaticality ofthe variants of the left-hand sentences in (10) featuring either inpreverbal position correlates with the grammaticality of theirnon-elliptical counterparts with coordination at the VP-level (cf.(15)), where the particle is RNR’ ed merely up to the VP.6

??

m

5 Han and Romero (2004:fn. 12) suggest, in connection with the RNR facts, thatspeakers might differ with regard to their tolerance of RNR of bare particles aboveIP, noting that some speakers accept the left-hand examples in (10) ‘with emphasis,

in certain discourse contexts’. They do not spell out these contexts; but it seems clearthat some of the factors that come to the rescue have little to do with the RNR’edparticle itself: thus, (ib) seems appreciably better than (ia) (which repeats the left-

hand sentence in (10b)) but both the weight and height of the particle up are plainlythe same in both sentences.

(i)a. ??Either they locked you or (they locked) me up.

b. (?)Either they locked a Romanian or (they locked) an Albanian up.6 Surprisingly, Schwarz (1999:360) never actually gives the non-elliptical versions

of (15a–c); instead, he provides examples of the same type that involve RNR of a PPor an adverbial (e.g., This either pleased Bill or pleased Sue a lot). He notes that ‘the

judgments are subtle’ (see also fn. 7 below); and indeed, it seems that the status of thevariants of (15a–c) with the bracketed material included in them is less than over-whelming. But, if only for the sake of argument, I will follow Schwarz (and, in his

wake, Han and Romero 2004 as well) and take them to be good.


(14)a. ??Either this pissed Bill or (it pissed) Sue off.

b. ??Either they locked you or (they locked) me up.

c. ??Either he gulped one or (he gulped) two down.

(15)a. This either pissed Bill or ((?)pissed) Sue off.

b. They either locked you or ((?)locked) me up.

c. He either gulped one or ((?)gulped) two down.

But what he does not point out is that variants of the examples in (14)can be constructed which, while still involving RNR to a positionabove IP, do come out grammatical. Two such variants are given in(16), differing from the examples in (14) in either lacking eitheraltogether or featuring it in a position left-adjacent to the verb, just asin (15).7

(16)a. (?)This (either) pissed Bill or it pissed Sue off.

b. (?)They (either) locked you or they locked me up.

c. (?)He (either) gulped one or he gulped two down.

The examples in (16), instantiating the ‘either too low’ puzzle, haveroughly the same status as the non-elliptical versions of those in (15)(which, as I pointed out in fn. 6, is somewhat tenuous), but since in(16) we are evidently dealing with coordination at the level of IP(unless we are willing to countenance an ‘unbalanced coordination’approach to (16); see fn. 9 for a critical appraisal), the particles musthave been RNR’ed just as high in (16) as in Schwarz’ s own examplesin (14). The fact that (14) and (16) differ in acceptability then suggests

7 The sentences in (16) are certainly not the preferred vehicles for the communi-cation of the messages they are meant to convey, but the same is true for the non-

elliptical variants of (15). Throughout, one would normally prefer not to RNR theparticle and to place it to the immediate right of the verb: <Either> this <either>pissed off<either> Bill or (it pissed off) Sue. I will return to the general awkwardness

of the V-NP-Prt order with focal pitch-accent on the NP in section 3.

MARCEL DEN DIKKEN

that an account in terms of the ‘extent of RNR’ is unlikely to be onthe right track.8

Whatever the relative merits of the movement and ellipsis ap-proaches to L-either, however, a major problem for both types ofaccount is that they do not seem to have an obvious handle onSchwarz’s (1999:368) observation, exploited by Han and Romero(2004), that whether . . . or . . . behaves differently from either . . . or . . .when it comes to the facts in (10): (10a–c) are all fine with whether (cf.(17) below).

(17)a. Whether this pissed Bill or Sue off is unclear.

b. Whether they locked you or me up, I can’t recall.

c. Whether he gulped one or two down is immaterial.

Moreover, the movement and ellipsis analyses are specifically keyedtowards an account of the ‘either too high’ puzzle (1), and havenothing to offer by way of a solution to the ‘either too low’ puzzleinstantiated by the example in (2b). This is a major conundrumfor an ellipsis approach: there simply is nothing to elide in (2b).Schwarz (1999:fn. 2) acknowledges the puzzle but does not providean analysis. A movement approach would, in order to accommo-date (2b), either need to resort to downward movement or assumean account whereby either itself does not move but the subject ofthe first disjunct raises around either and out of the coordination via(non-ATB) movement. But downgrading movement is generallyproblematic, and though the non-ATB subject movement analysismight perhaps be able to take care of (2b) as it stands, Hendriks(2003: 27) points out correctly that the plot thickens for a minimalvariant of (2b) featuring a modal, such as (2b¢):

(2b¢) John will either eat rice or he will eat beans.

Here John will is not a constituent, hence cannot move as a unit, andmovement of John and will individually (to obscure IP-externallanding-sites, moreover) would leave us with two violations of the

8 I do not profess to know the answer to the question of what makes (14) so muchpoorer than (16) for many speakers. That (16) has gone unnoticed in the literature is

due to previous work’s virtually exclusive emphasis on the ‘either too high’ puzzle.


Coordinate Structure Constraint. Though these would not necessarilybe fatal (much depends, after all, on a proper understanding of theroots of the CSC), it would nonetheless seem to diminish the chancesof an adequate solution to the ‘either too low’ puzzle couched in amovement-based analysis.9 While I am certainly not denying thatmovement and ellipsis both play a role in the syntax of either . . . or . . .,

9 Larson (1985) himself presents an ‘unbalanced’ coordination approach to (2b),having the VP of laughed coordinated with the IP he cried . Though this violates the

Law of Coordination of Likes in the overt syntax, Larson nonetheless seeks topreserve this law by declaring it active at LF only, and by arguing that in examplesthat apparently violate it, such as (2b), the two syntactically ‘unbalanced’ disjunctsare always of the same semantic type – thanks primarily to the fact that the overt

subject of the second disjunct is coreferential with the subject of the VP that con-stitutes the first disjunct. But while it is true that (2b) and perhaps the bulk ofR-either cases, the subjects of the two disjuncts are coreferential (something which

Hendriks 2003:fn. 13 links to the topicality of the second subject), there is no generalcoreference constraint on examples of this structural type. Sentences like those in (i)(of which (ib–f) were taken from Hendriks 2001, 2003; she culled them from the

literature, the press and the web) refute Larson’s claim that in ‘unbalanced’ coor-dinations the two subjects must be coreferential. Thus, any attempt at representingthese kinds of sentences as ‘unbalanced’ coordinations in overt syntax in the hopes ofgetting them to satisfy the Law of Coordination of Likes at LF is doomed to failure.

This is not to say, however, that the Law of Coordination of Likes is unsalvageablein the face of examples of this type. I will show below that they can be represented ascases of IP-level coordination, the fact that either shows up ‘buried’ inside the first

disjunct not standing out as a problem for the present analysis of R-either.

(i)a. You either leave now or I’ll call the police.

b. Yet our invitation was either a complete hoax . . . or else we had good

reason to think that important issues might hang upon our journey.(A. Conan Doyle, Sherlock Holmes)

c. By the prefix, we are either put in possession of the subject of the

poem, or some hint . . . is thereby afforded, not included in the body ofthe piece, which, without the hint, is incomprehensible. (E.A. Poe,Selected Works)

d. ‘Within a matter of a day or two,’ Dr. Yeomans said, ‘the situation will

become far more clear, and it will either become a nonevent or someappropriate announcement will be made. . .’ (New York Times, 20March 1998)

e. ‘I believe,’ he said, ‘positions will either harden or there will be asettlement in the next fewmonths.’ (NewYork Times, 15 February 1998)

f. The attachment will either open automatically or a dialog box will

appear requesting that you either open the file from its current locationor save the file. (Yahoo! Mail – Online Support; http://help.yahoo.com/help/us/mail/read/read-17.html)

MARCEL DEN DIKKEN

it seems plain that an integrated approach to both puzzles in (1) and(2) cannot be formulated exclusively in terms of either movement orellipsis.

3. CONTRASTIVE FOCUS AND THE SURFACE DISTRIBUTION OF EITHER

Something that makes many of the examples discussed towards theend of the previous section (cf. (10), (14)–(16)) rather tenuous is thefact that it is generally rather difficult to read the noun phrasesandwiched between a verb and a particle as a focus (as Kayne 1998points out; cf. also Chomsky 1971 on ditransitive constructions).Thus, (18b), like the right-hand examples in (10), is slightly degradedfor many speakers; the effect is particularly noticeable in contexts inwhich the focus on the sandwiched noun phrase is to ‘project’ up tothe VP: the answer in (19b) is significantly worse than that in (19a).10

(18)a. What is he looking up? He’s looking up a LINGUISTICS

TERM.

b. What is he looking up? ?He’s looking a LINGUISTICS

TERM up.

(19)a. What is he doing? He’s looking up aLINGUISTICS TERM.

b. What is he doing? ??He’s looking a LINGUISTICS

TERM up.

This prompts a discussion of the role of focus in the distribution ofeither.

3.1. Focus Matters, but either is not a Focus Particle Like Only

In her brief but insightful discussion of vagy . . . vagy . . . disjunctions(the Hungarian counterpart to English either . . . or . . .), Liptak (2001:22–34) establishes that ‘vagy . . ., vagy . . . structures necessarily in-volve exclusive focus constituents . . . in each parallel clause’ (p. 31).Converging with Liptak’s conclusion, Hendriks (2001, 2003, 2004)

10 On focus projection, see fn. 16 below. Note that Selkirk (1984:215–216)presents some wide-focus contexts in which focal stress on the ‘inner’ object in

ditransitive constructions is appropriate.


independently recognises the importance of focus in the syntax ofEnglish either . . . or . . . and its Dutch equivalent. Hendriks assimi-lates either (and neither and both as well; see esp. her 2004 paper) tofocus particles such as only, too, and also, although she is not veryspecific with respect to the syntactic details of an analysis of either . . .or . . . that treats either and its ilk as focus particles (Johannessen2005, who bases herself on Hendriks, gives more structuraldetails). The insight that the syntactic distribution of either is tiedto focus is indeed a very important one. Nonetheless, I believe itwould be wrong to syntactically assimilate either to focus particleslike only (as Johannessen 2005 does perhaps most explicitly). Letme briefly highlight the two most significant obstacles to such anassimilation.

There are two respects in which the distributions of either andfocus particles such as only/even do not match. On the one hand, thedistribution of only is narrower than that of either. This manifestsitself in at least the following two ways. First, as Hoeksema andZwarts (1991), McCawley (1996) and Kayne (1998) note, only andeven often do not felicitously attach to the (non-quantificational)complement of a preposition – thus, though the status of (20a) issubject to speaker variation, (20b) is entirely impossible for allspeakers. For either, however, we find no effect of this sort: (21a) and(21b) are both well-formed.11

(20)a. %John spoke to only/even Bill.

a¢. *To only Bill have they spoken the truth.

*To even Bill they wouldn’t tell the truth.

(21)a. John spoke to either Bill or Sue.

b. To either Bill or Sue, I will give a copy of my book.

A second respect in which the distribution of focus particles such asonly is narrower than that of either is illustrated by the fact that only/even refuse to attach all the way up to the IP in a context in which

11 Johannessen (2005:426–427) claims that sentences such as (21) are just as pooras the ones in (20), which is contrary to fact (although in R-either contexts an effectof this sort does indeed manifest itself – see section 4.2.4 below for discussion). I

could add here that the contrast between even and either is replicated in the contextof CP-complements (cf. ??He said even that he liked Bush vs. He said either that heliked Bush or that he wanted to give him the push; thanks to Aniko Liptak, p.c., for

pointing this out).

MARCEL DEN DIKKEN

they are associated to the direct object or the verb (cf. (22), whichillustrates this for only only). Hendriks (2003:10) notes this, and putsit down to ‘certain idiosyncratic properties of focus particles’.

(22)a. <*Only> John <?only> will <only> read <only>CHAPTER 3 <only> (not chapter 4 as well).

b. <*Only> John <?only> will <only> READ <*only>chapter 3 (not learn it by heart).

c. <Only> JOHN <only> will read chapter 3.

But what Hendriks does not point out is that, on the other hand, thedistribution of focus particles like only is broader than that of eitheras well. This manifests itself in at least two ways. First, as theexamples in (22a) and (22c) already indicate, only can surface to theright of the focus (John read chapter 3 only), something that eithercannot mimic (*John read chapter 3 either or chapter 4). Andsecondly, only, when associated to a focus in an embedded CP, cansometimes be placed outside that CP, in the matrix clause, as in (23a);for either this is impossible (cf. (8c) and (23b)).

(23)a. John <only> said that he would <only> read <only>CHAPTER 3 (not chapter 4 as well).

b. John <??either> said that he would <either> read<either> CHAPTER 3 or CHAPTER 4.

The grammaticality of (23a) with only in the matrix clause is hardlysurprising from a comparative perspective: focus movement canproceed successive-cyclically through SpecCP, as it does in, for in-stance, the Hungarian example in (24) (from E. Kiss 2000:255).

(24) JANOSTOLi mondtak [CP ti hogy szeretnek [CP ti haJanos-from said-they that would.like-they ifajanlast hoznek ti]]recommendation-ACC brought-IIt is Janos that they said that they would like if I brought arecommendation from.

The fact that L-either does not work the way only does in (23a) nowindicates that it is not associated with the focus the way only is.


3.2. Either and its Relationship with the Contrastive Focus

It is nonetheless incontrovertibly the case that either is closely tied tothe focus. More specifically, according to Liptak (2001:31) either . . .or . . . disjunctions demand ‘exclusive focus constituents . . . in eachparallel clause’ – a requirement that she relates to the interpretationof either . . . or . . . disjunctions (as distinct from their counterpartslacking either ): particularly, to the fact that either . . . or . . . is alwaysexclusive (cf. Liptak 2001:24). Thus, while a sentence such as Mary istaller than Bill or Bob supports an interpretation in which Mary istaller than both boys, its counterpart Mary is taller than either Bill orBob does not: Mary in this case is taller than just one of the boys,exclusively. The exclusiveness of either . . . or . . . disjunctions is di-rectly correlated with the requirement that they contain exclusive (or,equivalently, exhaustive) foci. This goes along with Hendriks’(2003:13) conclusion that ‘either needs the presence of a contrastivefocus in its c-command domain’, contrastive foci being exclusive/exhaustive (cf. Kenesei 1986; E. Kiss 1998).

Liptak’s and Hendriks’ references to an exclusive/contrastive focussuggest that the generalisation concerning the distribution of eithershould be stated in terms of the interpretive focus, not the focally pitch-accented constituent. That this is correct is shown by the triplet in (25).

(25)a. <Either> John <either> will <either> read<either> CHAPTER 3 or ((he’ll) read) CHAPTER 4.

b. <Either> John <either> will <either> read<*either> CHAPTER 3 or (he’ll) prepare DINNER.

c. <Either> John <either> will <either> read<*either> CHAPTER 3 or Jones will FLUNK him.

Here phonological focus in the first disjunct is systematically on thedirect object (CHAPTER 3) but the interpretive scope of contrastivefocus is different in the three examples (as marked by the underlin-ing12): in (25a) it is narrowly confined to the object, in (25b)contrastive focus is on the VP, and in (25c) the entire clause seems to

12 I have chosen underlining rather than Jackendoff’s (1972) square-bracket

notation for the demarcation of the focused constituent in order to avoid confusionwith the demarcation of the disjuncts, for which I will use the square-bracket nota-tion standard in syntax. The size of the focused constituents and the size of the

disjuncts are independent of one another, as the discussion shows.

MARCEL DEN DIKKEN

be contrastively focused. What we find in (25) is that it matters, whenit comes to the placement of either, what the scope of contrastivefocus is, not where the focal pitch accent is located: though in allthree examples the latter is located on the direct object, either cannotbe placed directly on the object in (25b, c) because the scope ofcontrastive focus is wider than the direct object alone.

That either can be placed on either side of will even in theapparent IP-focus case in (25c) suggests that the contrastive focusin this particular case is actually the Chomskyan vP (cf. Chomsky1995: Chapter 4 and subsequent work), the projection that har-bours the VP and the base position of the subject. By assigning theanswer in (25c) the structure in (25c¢), we make this examplecompatible with the generalisation that either must c-command thecontrastive focus (which, in the case at hand, will be vP, containingthe trace of the raised subject).13

(25c¢) <Either> Johni <either> will <either> [vP ti read<*either> CHAPTER 3] or [IP Jonesj will [vP tj FLUNK him]]

That the generalisation covering the surface distribution of eithershould specifically make reference to the locus of contrastive focusand not to information focus (new vs. old information) is confirmedby the question-answer pairs in (26)–(28).

(26)Q: I thought John would read chapter 2, but somehow hedidn’t. What did he read instead?

A: <Either> he <either> read <either> CHAPTER 3or ((he) read) CHAPTER 4.

(27)Q: I thought John would read chapter 2, but somehow hedidn’t. What did he do instead?


13 A reviewer points out that a somewhat similar situation, with the base positionof the subject in vP once again playing a key role, is found in the domain of the focus

particle even (cf. (ia)); only does not behave like even in this respect (cf. (ib)).

(i)a. JOHNi would even [vP ti speak to him]. � Even JOHN would speak to

him.b. JOHNi would only [vP ti speak to him]. „Only JOHN would speak to

him.


(28) Q: I thought John would read chapter 2, but somehowhe didn’t. What happened instead?


Here the preamble to the question as well as the answer are con-sistently the same, but the question is designed to have informationfocus on the object, the VP or the clause, respectively (as marked bythe dotted underlining in the answers). Throughout, A is a suitableanswer with either in any of its three positions. Though informationfocus has projected up to VP in (27) and to IP in (28) to match thequestion, the locus of contrastive focus remains confined to the di-rect object in all cases (as the solid underlining shows). As a result,placing either between the verb and the direct object systematicallydelivers a well-formed output in the answer.

Not only is it inconsequential what the extent of informationfocus is, the size of the disjuncts (or the level at which coordina-tion takes place) does not matter either when it comes to theplacement possibilities of either. As the parentheses indicate, thefacts in (25)–(28) hold regardless of whether the second disjunct isjust the object, the VP, or a complete clause. The focus facts foreither placement thus turn out not to discriminate between the‘either too high’ and the ‘either too low’ puzzles. The homogeneousbehaviour of the examples in (25)–(28) irrespective of whether thematerial in parentheses is included in them or not strongly suggeststhat the ‘either too high’ and the ‘either too low’ puzzles shouldnot be conceived of as separate puzzles, and should receive aunified account. Contrastive focus must play a key role in thataccount.

3.3. C-commanding the Contrastive Focus is Neither Necessarynor Sufficient

It is insufficient, however, to require of either (as does Hendriks2003:13) that it c-command ‘a contrastive focus’. While taking care of(29a–d), it leaves the ‘paired focus’ construction in (29e) by thewayside: all tokens of either in (29) have a contrastive focus in theirc-command domain, yet all variants of (29e) that have either in anyposition to the right of the first member of the pair of foci (JOHN) areungrammatical.

MARCEL DEN DIKKEN

(29)a. <Either> John <either> will <either> read<either> CHAPTER 3 or (he’ll read) CHAPTER 4.

b. <Either> John <either> will <either> READ

<*either> chapter 3 or (he’ll) DESTROY it.

c. <Either> John <either> WILL <*either> read<*either> chapter 3 or he WON’T (do so).

d. <Either> JOHN <*either> will <*either> read<*either> chapter 3 or MARY (will).

e. <Either> JOHN <*either> will <*either> read<*either> CHAPTER 3 or MARY CHAPTER 4.

We could of course modify the claim and require, more specifically,that either must c-command all contrastive foci in a given sentence.But this revised statement, while accurate for all of the examples in(25)–(29) and for a good number of other cases as well, does notachieve full descriptive adequacy either. In (30) (an expanded versionof (8), with solid underlining again demarcating the contrastivelyfocused constituent), all tokens of either meet that requirement, yetonly some pass muster: in (30b, c), either cannot be placed in thematrix clause, and many speakers also resist placement of either infront of the complementiser.

(30)a. <Either> he <either> wanted to <either> eat<either> RICE or BEANS.

b. <??Either> he <??either> wanted <%either> for you<either> to <either> eat <either> RICE or BEANS.

c. <??Either> he <??either> said <%either> that<either> he <either> would <either> eat <either>

RICE or BEANS.

While (29e) and (30b, c) instantiate cases where either is excludedfrom occurring in many locations from which it does in fact c-com-mand a contrastive focus, we also find cases in which, conversely,either legitimately occurs in positions that do not c-command a


contrastive focus at all.14 The question–answer pair in (31) (providedto me by an anonymous reviewer) is a case in point: here the contrast(between the question and the answer: (31A) would not work out ofthe blue) is external to the disjunction, but either immediately pre-cedes the first disjunct.

(31)Q: Did John say that he had either FRIED it or BAKED it?A: No! John DENIED that he had either fried it or baked it.

While the placement of either in the question is in keeping with arequirement to the effect that either must c-command a contrastivefocus, the fact that contrast in the answer is at the level of the matrixVP, which includes the disjunction as well as either, suggests thateither does not have to c-command the focus.

Not only does either fail to c-command the contrastive focus in(31A), it actually resists being placed in a position c-commanding thecontrastive focus (cf. (31A¢)). And while either is variably placed in asentence such as John said <%either> that <either> he <either>had <either> FRIED it or BAKED it (cf. (30c)), it seems that either in(31A) is fixed in the position immediately preceding the embeddedmain verb: the variants given in (31A¢¢) seem very awkward.

(31)A¢: No! <*Either> John <*either> DENIED that he hadfried it or baked it.

A¢¢: No! John DENIED <*either> that <*either> he<*either> had fried it or baked it.

In contexts of the type in (31A), therefore, it appears that either’sdistribution is entirely insensitive to the locus of contrastive focus,and is subject solely to the requirement that either be placed on thefirst disjunct.

14 A particularly peculiar case that meets the text description is Johannessen’s(2005:430) Norwegian R-either example in (i). The analysis of the distribution of

either offered in what follows does not directly accommodate (i), which does notseem to have a grammatical counterpart in English: *Peter WALKED either to the tramor he took the BUS is very awkward. The fact that the English verb does not move

whereas (i) is a Verb Second construction, featuring movement of the focused verbleaving a trace in a position c-commanded by enten, is presumably relevant; but howto make hay of this fact is unclear to me at this point (cf. also (37)–(38)).

(i) Per GIKK enten til trikken eller han tok BUSSEN. (Norwegian)Per walked either to tram-the or he took bus-the

MARCEL DEN DIKKEN

3.4. Either Placement: Contrastive Focus, and the ‘h-path’

To account for the constellation of facts canvassed in the foregoing,we need a theory of the distribution of either that will allow either tobe placed directly on the first disjunct (as, for instance, in (31A)) butwill also grant either the liberty to be embedded inside the first dis-junct (as in R-either cases) or to be outside the disjunction altogether(L-either). More microscopically, R-either must be subject to therequirement that it ‘float’ no further down than the first contrastivefocus (this is what remains of Hendriks’ c-command requirement),and L-either must be subject to restriction that it should not ‘runaway’ too far (recall (30b, c)). The descriptive generalisation in (32)ensures precisely this.15

(32) Either is a phrasal constituent in construction witha. the first disjunct, attaching to it; or

b. the first contrastive focus, attaching to

(i) the contrastive focus itself, or

(ii) a phrasal node on the h-path projected from the firstcontrastive focus.

15 The locution ‘in construction with’ is used non-technically in (32), though it isperfectly compatible with Klima’s (1964) technical notion (the predecessor of‘c-command’); the technical restrictions on either’s being ‘in construction with’ its

associate are defined for each individual case in (32a) and (32b.i/ii).An interesting question is whether either might be attached to the second disjunct

as well. The fact of the matter is that this is not the case (cf. (ia,b)). (32) rules this out

by stipulation, by making reference to the first disjunct and the first contrastive focusof a string of contrastive foci. The discussion in section 4.2.4, on or and its privilegedrelationship with the head ‘J’ of the coordination structure, will likely subsume theungrammaticality of both versions of (i) under the umbrella of minimality effects

(with (ia) then being a case where the desired Agree relationship between J and or isblocked by the intervening either, while (ib) would instantiate a failed attempt atmoving or across either). The inclusion of the modifier ‘first’ in the generalisation in

(32a) may ultimately be dispensable, therefore. I will keep it in for the sake ofconvenience.

(i)a. *He said that he would eat RICE either or BEANS.

b. *He said that he would eat RICE or either BEANS.


The notion of ‘h-path’ invoked in (32b.ii) is defined as in (33).16

(33)a. A h-path is a sequence of nodes such that each node ish-linked to the next higher node on the main projectionline.

b. a is h-linked to b iff a or its head assigns a h-role to b orreceives a h-role from b.

Before proceeding to an analysis of either placement from the per-spective of (32) and (33), let me address the way the h-path is com-puted. Take, for instance, a biclausal structure in which the object ofthe embedded verb is the contrastive focus. The verb selecting theobject h-marks its object, hence VP is h-linked to the object, by (33b).VP is itself the complement of I, which h-marks its VP complement –Selkirk (1984:209) made an early suggestion to this effect, andChomsky (1986:20) presented an explicit argument based on VPtopicalisation out of a wh-island: the fact that such extraction(illustrated in (34)) gives rise to only a mildly degraded resultindicates that the ECP is apparently being satisfied, which must meanin turn (given that antecedent-government of the VP-trace is

16 The notion ‘h-path’ and the way it is defined hark back to Kayne’s (1984)

theory of paths and connectedness. It also bears a resemblance to the Selkirkiannotion of ‘focus projection’ (cf. Selkirk 1995; Rochemont 1986). Although in earlierversions of this article, I sought to state the generalisation governing the (il)legitimate

L-either and R-either patterns in terms of Selkirk’s (1984) Basic Focus Rule in (ia)and, more importantly, her Phrasal Focus Rule in (ib), it has become clear to me thatactually formulating it in such terms is not feasible. Perhaps most devastating for afocus projection approach is E. Kiss’s (1998) demonstration of the fact that, while

information focus can project, contrastive focus (which she refers to as identifica-tional focus) cannot. If contrastive focus did indeed project up to the phrasal node onwhich either is placed and if L-either was sensitive to focus projection, the variable

placement of either should go hand in hand with variable scope for the contrastivefocus, quod non: regardless of where either is placed in a sentence such as (30a), thescope of contrastive focus is confined to the direct object.

(i)a. Basic Focus Rule (cf. Selkirk 1984:207)A constituent to which a pitch accent is assigned is a focus.

b. Phrasal Focus RuleA constituent may be a focus if (i) or (ii) (or both) is true:

(i) The constituent that is its head is a focus.

(ii) A constituent contained within it that is an internal argumentof the head is a focus.

MARCEL DEN DIKKEN

unattainable in this particular context) that I manages to h-govern itsVP complement.

(34) ?[VP Fix the car]i, I wonder [CP whether [IP he will ti]]

With I h-marking VP, it then follows that IP is h-linked to VP; andwith VP being h-linked to the object, we now have a h-path leading allthe way from the contrastively focused object to the IP. The h-pathcannot be extended further up, however: C does not h-mark IP, nordoes IP assign a h-role to C.

The significance of the fact that (33b) is stated in ‘in two direc-tions’, as it were, can be seen in contexts in which the container of thecontrastive focus is not h-marked by the head of the next higher nodebut instead assigns a h-role to (the head of) that node. Such is thecase, for instance, in sentences of the type in (35), where the con-trastive focus is the attributive AP, whose external h-role is assignedto the head noun (or, in the sense of Higginbotham 1985, this role is‘h-identified’ with the external h-role of the noun). In this way, the NPof the head noun is h-linked to the contrastively focused AP, so that ah-path can be constructed from the AP to the NP; that h-path extendsfurther up to the DP (on the assumption that the head noun’sexternal h-role is assigned to D, a la Williams 1981) and, thanks tothe thematic relationship between the DP and the verb, and betweenthe VP and I, all the way up to IP. This accounts for the fact thateither can be placed in any of the positions indicated in (35).17

17 Note that Selkirk’s (1984, 1995) focus rules (cf. fn. 16) will not get focus onBLUE to project to the complex noun phrase (as it should in the question–answer pair

in (i), where information focus rests on the entire direct object): the adjective isneither the head of the complex noun phrase (unless, perhaps, one adopts an Abney1987 type perspective on attributive modifiers) nor an argument of its head. In the

light of this and other examples, Schwarzschild (1999:167) comes to the conclusionthat the rules of focus projection are ‘simply wrong’; see also Buring (2002) andSelkirk (2003).

(i) Q: Nowadays John drives Mary’s red convertible. What did he drive

before?

A: He drove her BLUE convertible.

That either cannot actually be placed directly on the contrastively focused AP (see(iia)) is part of a broader generalisation (cf. (iib–d)) that I will not have space toaddress in detail in this paper. Suffice it to make a few brief remarks on the matter.

Hendriks (2003) brings up the ungrammaticality of things such as *right either ABOVE

or BENEATH that little chest to support her claim that either-disjunction is impossible


(35) <Either> he <either> drove <either> [DP his [NP

[AP BLUE] [NP car]]] or (he drove) his GREEN one.

Similarly, in (36), the fact that either can be legitimately placed on theVP, the I¢ or the IP even though contrastive focus is confined to themanner or temporal adverbial can be taken care of by (32) and (33)on the assumption that adverbial modifiers are predicated of the VPsthey modify, by which a h-link is established between the adverbialmodifier and the VP, with the h-path subsequently extending furtherup to IP. Naturally, placing either between the verb and the directobject in these examples is ungrammatical: adjunction of either to thedirect object violates (32).

(36)a. <Either> John <either> will <either> solve<*either> the problem <either> QUICKLY or (hewill solve the problem) THOROUGHLY.

(Footnote 17 Continued).below the maximal projection level, citing Neijt (1979). But on standard assump-

tions, what follows his in (iia,b), right in (iic) and carefully in (iid) is a maximalprojection. The ‘anti-locality’ effects on either-placement thus presumably have adifferent source. What we may be dealing with in these examples is a ban on

attachment of either directly to the lexical projection of the adjective, noun, prep-osition or verb. The empirical generalisation that either cannot be attached to a lexicalprojection can presumably be reduced to a general restriction imposed on lexical pro-jections: syntactic operations that duplicate a category label cannot apply to a lexical

projection. (This restriction will follow as a matter of course if lexical heads (and hencelexicalprojectionsaswell) donothaveacategory label (categorymembership thenbeingdeduced from the functional environment within which the lexical projection is

embedded), aswas suggested inMarantz (1997)and in subsequentworkdonewithin theframeworkofDistributedMorphology.)Analternative strategy that isworth exploringfurther in future work, in light of the discussion in section 4.2.4, would be to derive the

ungrammaticalityof therelevantvariantsof (ii) fromrestrictionsonor -movementtotheleft periphery of the second disjunct.

(ii)a. <Either> he <either> drove <either> his <*either> BLUE

car or ((he) drove) his GREEN one.

b. <Either> he <either> drove <either> his <*either>CONVERTIBLE or ((he) drove) his SUV.

c. <Either> he <either> put the ball <either> right <*either>

ON the box or ((he) put it) IN it.

d. <Either> he <either> carefully <*either> DISCUSSED the factsor (he) (slily) AVOIDED them.

MARCEL DEN DIKKEN

b. <Either> John <either> will <either> solve<*either> the problem <either> TOMORROW or(he will solve the problem) NEXT WEEK.

The facts in (35) and (36) show that h-paths can be projected fromconstituents that are syntactically adjuncts: whenever these constit-uents are predicated of the constituent to which they are adjoined,they are h-linked to their hosts, by (33b). The use, in the definition ofh-path in (33a), of the notion ‘h-linking’ (a symmetrical notion dis-tinct from ‘h-marking’, and perhaps comparable to Emonds’ 1985:78‘h-relatedness’) ensures that the account does not predict that con-texts in which either cannot be associated with a contrastive focus at adistance are precisely the familiar CED configurations, and hencedoes not reduce the analysis to a movement account a la Larson(1985). This is a desirable result.

The h-path is a syntactic notion, insensitive in principle to such PFproperties as the locus of focal pitch-accent. But in the specific con-text of the surface distribution of either, the locus of focal stress is ofthe utmost importance: L-either is regulated by a conjunction ofsyntactic and phonological conditions (the h-path and the placementof contrastive focal stress, respectively). The significance of thisconjunction of conditions is particularly evident in cases in which thecontrastive focus is ex situ, surfacing in a position that is not itself onany h-path but linked to a trace (or silent copy) that is. The examplesin (37b) and (38b) instantiate such cases.

(37)a. <Either> John <either> will <either> donate hisMONEY to the CHURCH, or (he will donate) his BOOKS tothe LIBRARY.

b. <Either> his MONEY, <*either> John <*either> will<*either> donate t to the CHURCH, or his BOOKS (he willdonate) to the LIBRARY.

(38)a. <Either> John <either> will <either> only his MONEY

or only his BOOKS.

b. <Either> only his MONEY will <*either> John<*either> donate t or only his BOOKS.


Plainly, a h-path leading up to IP is constructible in (37b) and(38b) from the base position of the fronted object. But apparently,the fact that a h-path can be so constructed is not sufficient toallow either to surface on the nodes on that path: attaching eitherto the projections of V and I is out of the question in these sen-tences, while it is perfectly well-formed in the a-examples. It wouldbe difficult to deny that a h-path can be built from the object’sbase position in the b-sentences: making the construction of a h-path sensitive to the presence of overt material at the foot of thepath is an indefensible move (because thematic notions are nototherwise known to be sensitive to the overt/covert distinction, norwould one expect them to be). But to facilitate L-either, we needmore than a h-path: we also need a focal pitch-accent. Focal pitch-accents can of course be realised only on overt material, not ontraces – and it is this that makes L-either directly sensitive to thephysical location of the first contrastive focus. The first focal pitch-accent encountered in the examples in (37b) and (38b) is that onMONEY. No h-path is constructible from the surface position of thispitch-accented constituent (which is an A¢-position). As a result, L-either is strictly impossible in sentences in which the contrastivefocus is preposed. The fact that the preposed focus binds a trace in aposition from which a h-path can be projected is irrelevant: the trace,being covert, can bear no focal pitch-accent.18

18 In the examples just discussed, A¢-movement was applied to the contrastivefocus, with drastic consequences for L-either. A question now arises about sentencesin which the (first) contrastive focus has putatively undergone A-movement: as in thecase of A¢-movement, the moved focus would not be in a position from which a

h-path is constructible, hence L-either is expected to be ungrammatical. Thisexpectation is not borne out: in the examples in (i), either is successfully placedfurther up the tree from the contrastively focused constituent. In the case of (ia, b),

one might want to have recourse to a (secondary) thematic relationship between theECM-subject and the matrix verb (cf. the discussion of ‘L-marking’ via Spec-Headagreement in Chomsky 1986); but while perhaps plausible for (ia) (though see

Hoekstra 1988 for a demonstration of the absence of a thematic relation between aperception verb and the subject of its small-clausal complement), a h-relationbetween consider and the ECM-subject would be a bit of a stretch, and for (ic) such a

strategy is definitely not viable. An integrated solution to the problem posed by(ia–c) from the perspective of the h-path approach to L-either will be forthcoming if

MARCEL DEN DIKKEN

With these background assumptions about the establishment ofh-paths in place, let me now proceed to laying out the analysis ofeither placement based on (32).

3.5. Either Placement and the h-path: Analysis

Let us turn, first of all, to the example in (30c), repeated here.

(30c) <??Either> he <??either> said <%either> that<either> he <either> would <either> eat<either> RICE or BEANS.

Since the verb eat h-marks the contrastively focused object, RICE , theembedded VP is on the object’s h-path.19 And by virtue of the h-relation between the I-head taking eat’s VP as its complement andthat VP, the h-path extends up to and including the embedded IP (cf.(39), which incorporates the structure of coordination constructions

(Footnote 18 Continued).passive and raising constructions have the subject merged in its surface position,hence it may attract its h-role (or aspectual feature, as in Manzini and Roussou’s2000 proposal) up the tree. I refer the reader to Manzini and Roussou (2000) for a

detailed defence of an analysis of ‘A-movement’ along these lines. (Note that thisdismissal of the standard movement-based account of raising and passive con-structions does not necessarily dispense with A-movement altogether; recall from the

discussion of (25c) in section 3.2 that some genuine A-movement dependencies, withconcomitant traces/copies, must exist on present assumptions. On the standardassumption that transitive, active vP is a strong phase, the need for a vP-internal

trace in (25c¢) can be derived if the h-role or aspectual feature that the subject inSpecIP needs to attract is invisible outside the vP phase (which will follow if therelevant property is a property of V).)

(i)a. <Either> he <either> saw <either> JOHN be arrested or BILL.

b. <Either> he <either> considers <either> JOHN to be the bestcandidate or BILL.

c. <Either> there <either> will <either> be <either> a MAN

arrested or a WOMAN.

19 The fact that the object of the verb eat is a disjunction, RICE or BEANS, is

immaterial. No matter what one’s assumptions are regarding the structure ofcoordination (see (3) and section 5, below, for a specific proposal, following standardassumptions in the current literature), one will always need to assume there to be athematic relationship between the verb and its coordinated objects to get the inter-

pretation and selectional restrictions right: a verb such as eat selects noun phrases,not conjunctions. I assume without discussion that the functional structure ofcoordinations is thematically and selectionally transparent.


in (3), discussed further in section 5; the circled nodes in (39) and therepresentations to follow in this section are on the h-path).

(39) He said [CP that [�IP he I=would [�VP eat [JP [ RICE ] [J [or[ BEANS ]]]]]]]

But (33b) stops the h-path in the embedded clause in (30c) at IP: CP isnot h-linked to IP, hence is not on the h-path. In (30c), therefore,either is predicted to be allowed to occur anywhere between theembedded complementiser and RICE: directly on the first disjunct orcontrastive focus, on the VP of eat, or on the IP above that, but nohigher. This is by and large accurate – but more needs to be saidabout speakers who can attach either to the embedded CP itself in(30c) (and also in (30b), which is parallel to (30c) in all relevantrespects).

I would like to argue that such speakers represent these sentencesas CP-coordination cases, with ellipsis in the second disjunct all theway down to the contrastive focus (cf. (40)). This is confirmed by thefact that speakers who accept (30c) with either to the immediate leftof that also accept (41A.ii) in answer to the question in (41Q), wherecontrastive focus is on the embedded object. For these speakers,therefore, (41A.i) is ambiguous between a parse in which the answeris structurally no larger than the noun phrase BEANS, on the one hand,and an alternative parse that structurally assimilates it to (41A.ii),with ellipsis down to BEANS.20 It is the latter strategy that is beingexploited in the analysis in (40).

20 Thanks to Jed Shahar and Erika Troseth for discussion of these data. Thespeakers accepting (30c) with either to the immediate left of that that I have con-

sulted also allow the answer in (41A.i) in reply to the question Did he say that hewould eat RICE or that he would eat BEANS?, with contrastive focus on the that-clause.If the size of the answer must minimally match the size of the contrastive focus,

(41A.i) in this context involves an elliptical that-clause of the type in (41A.ii); this willthen further confirm these speakers’ exploitation of ellipsis.Ellipsis is probably also at work in L-either cases such as (ia), pointed out to me

by Peter Culicover, differing minimally from (ib). For (ib), the account is entirely

straightforward: the during-PP is predicated of some (extended) projection of theverb, hence h-linked to that (extended) projection, and the h-path will extend furtherup to IP. The example in (ia) has as its second disjunct a constituent that is ostensibly

smaller than the during-adjunct (and as a result seems more delicate; the ‘?’ on thetoken of either preceding will is Culicover’s). I suggest that speakers acceptingL-either in (ia) represent it as a case of disjunction of a constituent larger than

during’s complement, with ellipsis down to the focused noun phrase.

MARCEL DEN DIKKEN

(40) [�IP He I [�VP said [JP [either [CP that he would eat RICE] [J [or

[CP that he’d eat BEANS]]]]]]]

(41)Q: Did he say that he would eat RICE or BEANS?

A.i: BEANS.

A.ii: %That he would eat BEANS.

Note that in (40), either is placed directly on the first disjunct eventhough the first disjunct is itself neither the contrastive focus (as isclear from its undergoing ellipsis; Wilder 1997 argues that ellipsisgoes all the way down to but never into the focus) nor on the h-path. This confirms our earlier conclusion (based on (31A),repeated below alongside its corresponding question) that thegrammar of either should allow placement of either on the firstdisjunct regardless of focus (cf. clause (32a)). Since either is not onthe h-path in (40), but attaches to a first disjunct that is neitheritself the contrastive focus nor on the h-path extending upwardsfrom the contrastive focus, either is allowed no further liberties; inparticular, it cannot be placed any higher up the tree. This ensuresthat speakers accepting (30c) with either to the immediate left ofthat still reject the versions of (30c) that have either placed furtherto the left. By the same token, it follows that either in the answerto (31Q) resists being placed in positions above or embedded insidethe first disjunct (cf. (31A¢, A¢¢)).

(31)Q: Did John say that he had either FRIED it or BAKED it?

A: No! John DENIED that he had either fried it or baked it.

(Footnote 20 Continued).

(i)a. John <?either> will <either> sneeze during the FIRST ACT or theINTERMISSION.

b. John <either> will <either> sneeze during the FIRST ACT or duringthe INTERMISSION.

The discussion of (40) and (i) makes it clear that the present account of eitherplacement does not eschew ellipsis as a matter of principle – there would be little

point in doing so, ellipsis being an incontrovertible ingredient of theory’s technicalmachinery. But unlike Schwarz (1999), my analysis resorts to ellipsis to a very limitedextent, assigning non-elliptical parses to the bulk of cases.


A¢: No! <*Either> John <*either> DENIED that he hadfried it or baked it.

A¢¢: No! John DENIED <*either> that <*either> he<*either> had fried it or baked it.

In (30c), contrastive focus is confined to the embedded object, asindicated by the underlining: a sentence such as (30c) can befelicitously used as a follow-up to a statement like John said thathe’d eat either of two things, with the presupposed set of alterna-tives (Rooth’s 1985 ‘p-set’) including {rice, beans}. As a follow-upto John said either of two things, with contrastive focus on theobject of the matrix verb said such that the p-set includes {thathe’d eat rice, that he’d eat beans}, sentences of the type in (30c) areusually not very felicitous. Instead, one would follow this up witha sentence such as (42).

(42) <Either> he <either> said <either> that<*either> he <*either> would <*either> eat<*either> RICE or (he said) that he’d eat BEANS.

Here, in contradistinction to what we see in (30c), attachment ofeither to the embedded clause becomes perfectly fine for all speakers,thanks to the fact that it is the embedded clause as a whole that servesas the contrastive focus of the complex sentence.21 In this case, eitheris also welcome to be inserted anywhere further up the matrix tree, onnodes that are on the h-path projected from the contrastively focusedCP (cf. (43)). But either now cannot be inserted anywhere inside theembedded clause; to do so would violate (32). This is true irrespectiveof the level at which disjunction takes place. In the variant of (42)that has that he’d eat BEANS directly following or, coordination isperformed at the level of the object of the matrix verb, as in (43a); inthe variant of (42) that has he said that he’d eat BEANS following or(instantiating the ‘either too low’ puzzle for all instances of either that

21 For Selkirk (1984, 1995), all phrasal focus is projected, via the Phrasal Focus

Rule (cf. fn. 16), from the pitch-accented element (here, the direct object of theembedded clause). With C not taking its IP-complement as its internal argument,focus should be unable to project from the lower verb’s object in (42) up to CP, or

from the lower verb up to the matrix VP in (44), below. Since there is no way oftracing the focus down to the pitch-accented element in keeping with the PhrasalFocus Rule, it would be impossible, by Selkirk’s logic, to take the embedded CP in

(42) and the matrix VP in (44) (of the first disjunct) to be the focus.

MARCEL DEN DIKKEN

do not precede the matrix subject in the first disjunct), coordination isat the level of the matrix IP, as illustrated in (43b). Regardless,however, the h-path runs from the object of said up to the matrix IP,and hence, regardless of the size of the disjuncts, either may beattached to the embedded CP, the matrix VP or the matrix IP, inconformity with (32), but not anywhere inside CP.

(43)a. [�IP He I [�VP said [JP [CP that he would eat RICE ] [J [or [CPthat he would eat BEANS ]]]]]]

b. [JP [�IP He I [�VP said [CP that he would eat RICE ]]] [J [or [IPhe said [CP that he would eat BEANS ]]]]]]

If contrastive focus is anchored even higher in the tree, at the level ofthe matrix VP, either may not even be placed on the embedded CP.This is shown in (44), which is analysed as in (45).

(44) <Either> he <either> said <*either> that <*either>he <*either> LOVED her or (he) INSULTED her.

(45)a. [�IP He [JP [VP said that he LOVED her ] [J [or [VP INSULTED

her ]]]]]

b. [JP [�IP He [VP said that he LOVED her ]] [J [or [IP he [VPINSULTED her ]]]]]

The natural interpretation of (44), as a follow-up to something likeJohn did either of two things, is such that a contrast is establishedbetween his saying something and his insulting her. It should besaid, however, that on a – considerably less natural – interpreta-tion in which a contrast is established at the level of the embeddedVP (following up on John said he did either of two things), (44) willbehave to all intents and purposes like (30c) (or like John said thathe LOVED her or (he) HATED her), hence on such a reading place-ment of either in the two positions inside the embedded clause in(44) would be legitimate while placement of either in the matrixclause is ungrammatical, with speakers varying once again on thequestion of whether either can immediately precede that in such acontext.

It should be plain that the account of either placement presentedhere also directly accommodates the facts in (6) and (7), above (from


Larson 1985). Larson’s observation that negation blocks L-either (cf.(5), repeated below as (46)) can be accounted for as well.

(46)a. (?)John didn’t eat either RICE or BEANS. (cf. (5))

b. ?? John either didn’t eat RICE or BEANS.

c. ?? Either John didn’t eat RICE or BEANS.

Either can be legitimately placed directly on the contrastively fo-cused constituent (RICE), as in (46a), but positioning either to theleft of the negation, as in (46b, c), results in a strongly degradedresult. This is because negation breaks the h-path leading up fromthe contrastive focus. With negation syntactically represented interms of a NegP, the fact that its head, Neg0, does not h-mark itscomplement (which I take to be VP, in a non-split IP system)prevents the h-path from extending beyond the complement ofNeg. Placement of either to the left of negation will then be ruledout, which results in (46b,c) being rejected as sentences in whichthe direct object is the contrastive focus, and disjunction is at thelevel of the direct object.

Interestingly, when disjunction is at the level of IP, as in (47), thejudgements flip: (47b,c) are now fully well-formed, and (47a) isungrammatical.22

(47)a. *John didn’t eat either RICE or he didn’t eat BEANS.

b. John either didn’t eat RICE or he didn’t eat BEANS.

c. Either John didn’t eat RICE or he didn’t eat BEANS.

22 That (46b,c) apparently do not (naturally) support an IP-coordination parse

with ellipsis down to the contrastive focus (cf. Schwarz’s 1999 ellipsis-based analysisof L-either) is presumably due to the fact that negation in the second disjunct is beingelided on such a parse. The fact that (46a) is not perfect may tie in with this: (46a) is

ambiguous, in principle, between a ‘neither’ reading (John ate neither rice nor beans)and a ‘one of the two’ reading (John didn’t eat rice or he didn’t eat beans); but thelatter reading, which corresponds to the ones that (46b,c) would legitimately support

if they (naturally) allowed an IP-coordination parse, is salient only in reply to neg-ative questions such asWhat didn’t John eat/do? orWhat is it that John didn’t eat/do?;and the ‘neither’ reading is more felicitously expressed with the aid of neither itself,

rather than with a combination of negation and either.

MARCEL DEN DIKKEN

The grammaticality of (47c) is of course straightforward: either hereis placed directly on the first disjunct, in accordance with (32a). Thesentences in (47a) and (47b) are both cases of R-either, the differencebetween the two being that in (47a) either has ‘floated rightward’beyond the sentential negation while in (47b) it stays to the left of thenegation. But R-either does not literally involve rightward (i.e.,downward) float of either: such would be an illegitimate case ofdowngrading movement. We cannot therefore hold the position ofeither vis-a-vis the sentential negation directly responsible for thedifference between (47a) and (47b). On my analysis of R-either,‘rightward floated’ either is attached to a node within the first disjunctthat either is itself the contrastive focus or finds itself on the h-pathprojected from the contrastive focus. Since the h-path projected fromthe focally pitch-accented direct object RICE does not extend to NegP(as we discovered in the previous paragraph), (47b) must be a case inwhich the contrastive focus is the NegP itself, with either placeddirectly on the contrastive focus. This is grammatical. But it is notimmediately apparent why (47a) should be ungrammatical: RICE,after all, is a possible contrastive focus (it is in (46a)), so why shouldeither not be welcome to attach to it? The analysis of either . . . or . . .disjunctions developed up to this point does not account for theungrammaticality of (47a). As we will discover in section 4, the causeof this hole in the account is our one-sided focus, up to this point, oneither as the root of all restrictions on either . . . or . . .: as it will turnout, the root of the problem with (47a) actually lies with or, not witheither.

In the present section, we have seen that analysing either as aphrasal category sensitive to the locus of contrastive focus and what Ihave called the ‘h-path’ gives us good mileage on L-either andR-either, without recourse to movement of either, and without ref-erence to either the size of the disjuncts or the scope of informationfocus. In the next section, we will find that for a comprehensivepicture of the syntax of either . . . or . . ., we also need to take or intoaccount – in particular, we need to analyse or, like either, as a phrasalcategory; unlike either, however, or is forced to surface at the left edgeof its disjunct, so if it is not base-generated there (but instead origi-nates somewhere inside the second disjunct), it must raise there in thecourse of the syntactic derivation. Movement of or is subject tosyntactic restrictions, and it is these syntactic restrictions that willultimately give us an answer to why (47a) is ungrammatical.


4. EITHER AND OR AS PHRASAL CATEGORIES

The fact that either can occur in positions on the first disjunct, thecontrastive focus or along the h-path projected from the contrastivefocus suggests that either occupies a PHRASAL position.23 Beingphrasal, it should distribute like things phrasal. In this section I willshow first of all (in section 4.1) that this gives us an account of theinteresting restrictions on either placement to the immediate left of asubject, and I will subsequently show (in section 4.2) that thephrasality of either allows its negative and [+WH] incarnations,neither and whether, to undergo movement to positions beyond thosealong the contrastive focus’s h-path that they can be base-generatedin. After extending the phrasal approach to (n)or in section 4.3, I willthen lay out in section 4.4 the analysis of coordination constructionsthat the discussion gives rise to. Finally, I present some furthersupport for the phrasal approach to either and or in section 4.5, fromlocality effects that would otherwise go unexplained.

4.1. Either as a Phrasal Adjunct, and the Restrictions on its Placementin Pre-subject Position

The conclusion that either is a phrasal category that attaches tomaximal projections is of course compatible with our earlier discus-sion of ‘either-float’. In the present subsection, I will argue that it isfurther confirmed by the examples in (48)–(53).

(48)a. John considers <*either> the president <either> aFOOL Or a GENIUS.

b. John considers <?either> the PRESIDENT <*either>a FOOL or his WIFE a GENIUS.

c. John considers <either> the PRESIDENT <*either>a fool or his WIFE.

(49)a. John put <*either> the book <either> on<?either> the SHELF or (on) the TABLE.

23 See also Johannessen (1998, 2005), who draws the same conclusion. She pre-

sents evidence from a variety of Germanic Verb Second languages to show thateither, when initial in a finite disjunct, can and in some languages must trigger VerbSecond (cf. e.g. the first disjuncts of German (70a,b), below). On (n)or in the second

disjunct triggering inversion, see section 4.2.3 and fn. 31.

MARCEL DEN DIKKEN

b. John put <?either> the BOOK <*either> on theSHELF or the RECORD on the TABLE.

c. John put <either> the BOOK <*either> on the shelfor the RECORD.

(50)a. John gave <*either> the book <either> to<?either> MARY or (to) SUE.

b. John gave <?either> a BOOK <*either> to MARY

or a RECORD to SUE.

c. John gave <either> a BOOK <*either> to Mary or aRECORD.

(51)a. John gave <*either>Mary <either> a BOOK or aRECORD.

b. John gave <?either> MARY <*either> a BOOK orSUE a RECORD.

c. John gave <either> MARY or SUE <*either> a book.

(52)a. John considers <*either> the president <either>to <either> be <either> a FOOL or a GENIUS.

b. John considers <?either> the PRESIDENT to be a FOOL

or his WIFE a GENIUS.

c. John considers <either> the PRESIDENT to be a foolor his WIFE.

(53)a. John wants <*either> Mary <either> to <either> eat<either> RICE or BEANS.

b. John wants <?either> MARY to eat RICE or SUE BEANS.

c. John wants <either> MARY to eat rice or SUE.

Throughout these examples, either c-commands the contrastive focus;nonetheless it cannot be placed to the immediate left of the subject ofthe small or ECM-infinitival clause in the a-sentences. The general-isation covering the examples in (48)–(53), which I will show can bemade to follow from the account of the surface distribution of eitherdeveloped here, is that either cannot be inserted to the immediate leftof the subject of a small-clausal or ECM-infinitival complement


unless that subject is itself the contrastive focus or the higher memberof a pair of contrastive foci.24

Why would this generalisation hold? With either being a phrasalcategory, a word order in which either immediately precedes thesubject of a small or infinitival clause could in principle be analysed ineither of two ways: either either is attached to the subject itself, or it isadjoined to the small or infinitival clause as a whole. The latter isprohibited: it is impossible to adjoin any material to a small orinfinitival clause in complement position (cf. (54)), something whichhas traditionally been blamed on Case Theory (Stowell’s 1981 adja-cency condition on structural Case assignment).

(54)a. John <basically> considers <*basically> the president<basically> a fool.

b. John <carefully> put <*carefully> thebook<carefully> on the table.

c. John <secretly> gave <*secretly> the book<secretly> to Mary.

24 See Kayne (1984), Hoekstra (1988), Den Dikken (1995) and references cited

there for small clause approaches to resultative, prepositional dative and doubleobject constructions. I should add here that, throughout (48)–(53), either can felici-tously be placed to the left of the verb or the matrix subject, as expected in light of the

discussion of the h-path in section 3. Note that such placement of either is possibleeven in the case of want-constructions – something which, in light of the fact that (8b)is poor with either in the matrix clause (cf. <??Either> John <??either> wanted

<%either> for you to eat <either> RICE or BEANS), suggests that the ECM com-plement of verbs like want is not a covert for-to infinitive (contra e.g. Kayne 1984): ah-path can be projected from the infinitive’s object into the matrix clause. I should

also note that the paradigms in (48)–(53) can be reproduced in their entirety forR-either (e.g., John considers <*either> the president <either> a FOOL or he con-siders him a GENIUS). This once again confirms that R-either should not be given some‘special’ solution but should be part and parcel of the general analysis of either

placement. That R-either is possible (though somewhat marginal) in want-ECMconstructions as well (cf. John wants Mary to<?either> eat <?either>RICE or hewants her to eat BEANS) while such is entirely impossible in for-to infinitives (cf. John

would prefer for Mary to <*either> eat <*either> RICE or he would prefer for her toeat BEANS) is an additional indication that the ECM complement of want is not a CP.By the same logic, the grammaticality of both L-either and R-either in control con-

texts (cf. <Either> John<either> wants to <either> eat < either> RICE or (hewants to eat) BEANS) suggests that control infinitives are smaller than CP as well. I willcontent myself here with pointing out these consequences without pursuing them

further.

MARCEL DEN DIKKEN

d. John <secretly> gave <*secretly> Mary<secretly> the book.

e. John <secretly> considers <*secretly>the president to be a fool.

f. John <passionately> wants <*passionately> Mary toeat rice.

So the generalisation that either cannot be inserted to the imme-diate left of the subject of a small-clause or ECM-infinitivalcomplement unless its subject is itself the contrastive focus or thehigher member of a pair of contrastive foci will follow if we canensure that either can only be adjoined to the subject of the smallclause if that subject is focused; if, by contrast, the subject is not infocus, either cannot be inserted between the verb and the ECM-subject, on pain of a violation of the same general ban that alsomanifests itself in the examples in (54). The representations in (55)sum this up.

(55)a. John gave [SC (*either) [SC [(*either) a book] [PP toMARY]]] . . .

b. John gave [SC (*either) [SC [(?either) a BOOK] [PP to MARY]]]. . .

c. John gave [SC (*either) [SC [(either) a BOOK] [PP to Mary]]]. . .

That attaching either to the subject of the small clause is impossiblein (55a) is because this subject is not the first disjunct, the firstcontrastive focus or a node on the h-path projected from thefirst contrastive focus (i.e., MARY). That the SC-subject is not the firstdisjunct or the first contrastive focus is plain; it may not be imme-diately obvious, however, that it is not on the h-path projected fromMARY either. The PP headed by to is certainly h-linked to the con-trastive focus, hence on the h-path; and the SC-subject, receiving anexternal h-role from the to-PP, is h-linked to the to-PP (cf. (33b)). Butthat does not make the SC-subject be on the h-path projected fromthe contrastive focus: paths (as in Kayne’s 1984 theory) are defined,in (33a), in such a way that all nodes on them other than the lowestone (i.e., the one from which the h-path is projected) must be on themain projection line.


For the c-examples in (48)–(53), where the embedded subject isthe sole contrastive focus of the sentence, base-generation of eitheron the subject (as in (55c)) is of course perfectly straightforward.Interestingly, however, we are led to base-generate either there inpaired focus constructions of the b-type as well: the variants of theb-examples in (48)–(53) with either to the immediate left of thesubject result not from attachment of either to the small orinfinitival clause as a whole but rather from attachment of either tothe first member of the pair of focused constituents (as depicted in(55b)).

The treatment of either as a phrasal adjunct and the constituencyof either and focused subject (whether it be the sole focus or the firstmember of a pair of foci) can both be confirmed by blocking effects.Thus, while in (56a), either can be placed in any of three positions,wh-fronting of to whom with concomitant subject-auxiliary inversion,as in (56b), is successful if either is placed directly on the focuseddirect object or on the VP, but not if either is inserted to the left of thesubject (where it is free to be placed if no wh-fronting plus concom-itant I-to-C movement takes place).

(56)a. <Either> John <either> gave <either> a BOOK toMary or a RECORD.<Either> John <either> gave <either> a BOOK or aRECORD to Mary.

b. To whom did <*either> John <either> give <either>a BOOK or a RECORD?

In this respect, either behaves precisely like phrasal adverbs such asprobably, which likewise exert a blocking effect on inversion, as seenin (57).25

(57) To whom did <*probably> John <probably> give aBOOK?

25 That the blocking effect in (56b) and (57) is one involving subject-aux inversionand not wh-fronting is evident from the fact that in embedded questions, the effect

disappears: thus, both I’d like to know to whom (either) John gave a BOOK or aRECORD and I’d like to know to whom (probably) John gave a book are grammaticalwith the bracketed material included. (The roots of the blocking effect seen in (56b)

and (57) are not very well understood; see Kayne 1984:Chapter 10 for an interestingperspective.) On root and embedded yes/no-questions, pre-subject either and prob-ably diverge, which is something I will address in the context of intervention effects in

section 4.2.2, below; see the main-text discussion of (67) and also fn. 29.

MARCEL DEN DIKKEN

But either IS allowed to linearly intervene between the raised auxiliaryand the subject in (58a,b), where the subject is focused (either by itselfalone or in a paired focus construction). This mimics the absence of ablocking effect on inversion when an adverb is a subconstituent of thesubject, as in (59b), where often modifies the gerund in subject po-sition and thus appears embedded within the matrix subject (unlike inthe minimally contrasting case in (59a), where often is construed as amodifier of the matrix predication).

(58)a. To whom did [(either) JOHN] give a book or BILL?

b. To whom did [(either) JOHN] give a BOOK or BILL a RECORD?

(59)a. To whom does (*often) [working on focus] seem appealing?

b. To whom does [(often) working on focus] seem appealing?

The discussion in this section has shown that the restrictions on eitherplacement seen in (48)–(53) and (56)–(59) provide substantial evidenceto support the view that either is a phrasal category – an adjunctattaching to eligible maximal projections. In the next section, I willfurther support the phrasality of either, as well as neither, whether, orand nor, on the basis of restrictions on phrasal movement.

4.2. Phrasal Movement

4.2.1. Either vs. WhetherL-either (i.e., either placed in a position higher in the tree than thoseleft-adjacent to the first disjunct or contrastive focus) is not readilyaccepted by all speakers: speakers disagree in their appreciation of thevariants of (29) (repeated below) with either ‘running away’ from thecontrastive focus.

(29)a. <Either> John <either> will <either> read<either> CHAPTER 3 or (he’ll read) CHAPTER 4.

b. <Either> John <either> will <either> READ

<*either> chapter 3 or (he’ll) DESTROY it.

c. <Either> John <either> WILL <*either> read<*either> chapter 3 or he WON’T.

d. <Either> JOHN <*either> will <*either> read<*either> chapter 3 or MARY (will).

e. <Either> JOHN <*either> will <*either> read<*either> CHAPTER 3 or MARY CHAPTER 4.


Even those who do not like L-either do, however, happily acceptwhether (the [+WH] counterpart of either; cf. Larson 1985; Jespersen1961: Vol. II, p. 200; Vol. III, p. 43; Vol. V, p. 384) ‘running away’from the focus: the whether examples in (60) are perfectly fine for allspeakers.

(60)a. I would like to know whether John will read CHAPTER 3 orCHAPTER 4.

b. I would like to know whether John will READ chapter 3or TEAR it UP.

c. I would like to know whether John WILL read chapter 3 orhe WON’T.

d. I would like to know whether JOHN will read chapter 3or MARY.

e. I would like to know whether JOHN will read CHAPTER 3 orMARY CHAPTER 4.

Of course, there is an explicit trigger for leftward movement ofwhether: the fact that it is marked [+WH] forces it to be in SpecCP,regardless of where it originated and regardless of the question(which speakers apparently react to differently, perhaps for stylisticreasons26) of whether either is welcome only on the contrastive focusitself or along the h-path projected from the contrastive focus.

The very fact that either does not ‘run away’ freely and issometimes actually prevented from ‘running away’ (as in the RightNode Raising cases discovered by Schwarz 1999; see section 2),while whether is welcome in all contexts in which L-either is not, asis seen in the contrast between Schwarz’s (61) and their whethercounterparts in (62) (cf. Schwarz 1999: 368; Han and Romero2004), suggests that either can only occur in positions in which itcan be base-generated: it is not allowed to stray further from homevia movement in syntax.

26 I say stylistic because I have not found any clear grammatical restrictions atwork here in English. There certainly are languages in which the equivalent of eitheris categorically prevented from floating (cf. e.g., French ou ‘either/or’ and ni ‘neither/

nor’), but as far as I know, no variety of English is in that category.

MARCEL DEN DIKKEN

(61)a. ??Either this pissed Bill or Sue off. (cf. (10))

b. ??Either they locked you or me up.

c. ??Either he gulped one or two down.

(62)a. Whether this pissed Bill or Sue off is unclear. (cf. (17))

b. Whether they locked you or me up, I can’t recall.

c. Whether he gulped one or two down is immaterial.

I therefore side with Han and Romero (2004) in prohibiting eitherfrom moving. They attribute this to a parallel between either andadverbial material, noting that the fact that adverbs such as often donot move by themselves while their [+WH] incarnations (how often)do is matched by the contrast between either and whether when itcomes to movement.

4.2.2. Phrasal Movement of Whether: Intervention Effects27

That whether is a phrase and moves as a phrase is clear from the factthat it must occur in a position which is known to be reserved forphrases (SpecCP) and which, moreover, is outside the h-path: thoughthe h-path extends from the complement of V up to the VP andfurther up to the IP, it is impossible for the h-path to be extendedfurther up to CP because IP is not h-marked by C. That whetherarrives in SpecCP via movement is confirmed in an interesting way bythe fact that its raising to this position is interfered with by ‘harmfulinterveners’ – either, in particular, blocks movement of whether acrossit. To set this up, consider first the example in (63). This example isacceptable in six surface guises, spelled out in (63a–f). The ones thatinterest me here in particular are those in (63e,f), both featuring twotokens of either, one for each disjunction.

(63) <Either> Mary <either> went <either> SWIMMING orDANCING, or she stayed at HOME.

a. Mary went SWIMMING or DANCING, or she stayed at HOME.

b. Either Mary went SWIMMING or DANCING, or she stayed atHOME.

27 The discussion in this section draws in part on Gullı (2003), which also dis-

cusses Italian and Calabrian in this connection.


c. Mary either went SWIMMING or DANCING, or she stayed atHOME.

d. Mary went either SWIMMING or DANCING, or she stayed atHOME.

e. Eitherl Mary either2 went SWIMMING or DANCING, or shestayed at HOME.

f. Either1 Mary went either2 SWIMMING or DANCING, or shestayed at HOME.

When we replace either1 in (63e,f) with its [+WH] counterpartwhether and try to raise it to SpecCP, the result is grammatical, asseen in (64a), but replacing either2 with whether and raising it acrosseither1 delivers an ungrammatical output (cf. (64b)); the ungram-matical example in (64b) can be rescued, though, by turning it into apaired focus construction, as in (64c).28

(64)a. I wonder whether Mary (either2) went (either2) SWIMMING

or DANCING.

b. I wonder whether (*either1) Mary went SWIMMING orDANCING.

c. I wonder whether (either1) MARY went SWIMMING or SUE

went DANCING.

That (64a) is unproblematic with either present is not particularlysurprising: the base position of whether (which here corresponds toeither1 in (63e,f)) is higher up the tree than the position occupied byeither2. The interesting thing is that (64a) becomes ungrammaticalwhen either is attached to the IP (as in (64b)), but that with either inexactly the same place in the linear string, (64c) is nonetheless okay.Taking (63e,f) as our model fromwhich to derive the sentences in (64a)and (64b), we can capture the ungrammaticality of (64b) with eitherincluded as an intervention effect: the position occupied by either1 (i.e.,an IP-adjoined position) c-commands and is of the same type as thebase position of whether (which in (64b) is the [+wH] counterpart ofeither2 in (63e,f)) – with whether crossing over either we then get an

28 To keep the sentences relatively simple, I suppressed the additional disjunct orshe stayed at HOME (or or they stayed at HOME, in the case of (64c)). This is alegitimate move: whether-constructions in general do not need an or-phrase (for

reasons that are not entirely transparent but need not concern us here).

MARCEL DEN DIKKEN

ungrammatical result, as desired. The representation in (65) aims tobring this out for the variant of (64b) corresponding to (63e), withwhether originating in a position adjoined to the VP of went.

(65) [CP whether2 [IP either1 [IP Mary [VP t2 went [SWIMMING or DANCING]]]]]

That (64c) is grammatical must then mean that no intervention effectmanifests itself there. And this follows readily in the light of theaccount of paired focus constructions presented in section 4.1: suchconstructions have either forming a constituent with the first memberof the pair of foci (here, the subject). Hence, no intervention effectmanifests itself in the paired focus case in (64c): with either adjoinedto the subject, it is not going to interfere with movement of whether inany way whatsoever.

Note that the examples in (66) pattern exactly like those in (64).

(66)a. I wonder if Mary (either2) went (either2) SWIMMING orDANCING.

b. I wonder if (*eitherl) Mary went SWIMMING or DANCING.

c. I wonder if (either1) MARY went SWIMMING or SUE wentDANCING.

If the analysis of the latter outlined in the previous paragraph is onthe right track, this strongly confirms the hypothesis that in embed-ded if-questions there is null operator movement to SpecCP. And thatanalysis then readily carries over to root yes/no-questions as well (cf.(67)).29

(67)a. <Either> you <either> ate <either> RICE or BEANS.

b. <*Either> did <*either> you <either> eat <either>RICE or BEANS?

·m

29 See also Han and Romero (2004). The ungrammaticality of *Did either you eatrice or beans? is thus taken care of independently of the ungrammaticality of (56b)above, with either placed between did and John. That this is a good result is clear

from the fact that, while the deviance of the relevant variant of (56b) vanishes inembedded contexts (cf. fn. 25 for illustration), that of (67b) with either to theimmediate left of the subject persists: I was wondering whether <*either> you

<either> ate <either> rice or beans.


Such an analysis is enhanced by the fact that in several languages,including Early Modern English (cf. Radford 1997:295; see (68) forillustration), an overt incarnation of this operator shows up inSpecCP: none other, in fact, than the element whether.

(68)a. Whether had you rather lead mine eyes or eye yourmaster’s heels? (Mrs Page,Merry Wives of Windsor, III.ii)

b. Whether dost thou profess thyself a knave or a fool?(Lafeu, All’s Well that Ends Well, IV.v)

4.2.3. Phrasal Movement of Neither and Nor: Negative InversionHaving established that the [+WH] incarnations of either, i.e.,whether and its null-operator counterpart, are phrasal categoriesundergoing phrasal movement, I now proceed to showing that for its[+NEG] incarnation, which is spelled neither, a similar story applies.In (69b) (cf. Culicover 1999) we find neither surfacing in a positionwhich is known to be restricted to phrases (the specifier position fromwhich Negative Inversion is triggered) and which clearly lies outsideIP, the largest node to which a h-path can be projected from anythinginside IP.

(69)a. Mary neither spends her vacations at the seashore nordoes she go to the mountains.

b. ?Neither does Mary spend her vacations at the seashore nordoes she go to the mountains.

c. *Neither Mary spends her vacations at the seashore norshe goes to the mountains.

The examples in (69) also show that nor, the [+NEG] counterpart ofor, exhibits phrasal behaviour as well: it, too, triggers NegativeInversion (obligatorily so), hence must occupy the IP-external A¢-specifier position from with Negative Inversion is accomplished.30

The case of German is interesting in this connection as well.Consider the pair in (70) (from Wesche 1995; see also Lechner 2000;Johannessen 2005:436).

30 Culicover (1999:164) contests the claim that Negative Inversion is triggeredfrom a specifier position, saying that one might �abandon the requirement that

negative inversion occurs only when what appears to the left of the inverted auxiliary

MARCEL DEN DIKKEN

(70)a. Entweder kocht Hans heute, oder Maria ruft deneither cooks Hans today or Maria calls thePizzaservice. (German)pizza serviceEither Hans cooks today or Maria calls the pizza service.

b. Weder kocht Hans heute, noch ruft Maria denneither cooks Hans today nor calls Maria thePizzaservice.pizza serviceNeither will Hans cook today nor will Maria call the pizzaservice.

(Footnote 30 Continued).is a phrase’, and going on to suggest that negative inversion can ‘simply’ be stated ‘interms of the linear sequence in the structure, respecting of course the organizationinto phrases’, as in (i).

(i) X[+AFFECTIVE] NP Vfin Y �X[+AFFECTIVE] Vfin NP Y

The reason why such a linear statement is untenable (even abstracting away fromthe awkward reference to ‘Vfin’: only finite auxiliaries undergo inversion, of course) isthat [+AFFECTIVE ] matrix verbs that happen to be linearly adjacent to the subject of

their embedded clause (thanks to complementiser deletion) never bring about Neg-ative Inversion – take the case of deny, illustrated in (ii).

(ii)a. Newcastle United footballer Laurent Robert has denied he wasinvolved in a scuffle with a local newspaper reporter.

a¢. *Newcastle United footballer Laurent Robert has denied was heinvolved in a scuffle. . .

b. Mr Vizard, a former lawyer, also denied he had engaged in insider

trading as a former Telstra director to invest in a computercompany Sausage Software.

b¢. *Mr Vizard, a former laywer, also denied had he engaged in insidertrading. . .

c. John Armitt denied he had taken on the worst job in Britain, sayinghe liked a ‘challenge’.

c¢. *John Armitt denied had he taken on the worst job in Britian. . .

So discarding Culicover’s Negative Inversion rule in (i), we are left with the textconclusion that the fact that neither and nor trigger Negative Inversion shows thatneither and nor are phrases that may sit in specifier positions.


Whereas oder ‘or’ does not trigger inversion after it in the seconddisjunct, noch ‘nor’ does. Placing the disjunction particle on the leftedge of the second disjunct does not, therefore, give rise to inversionas a matter of course: we do not get a generalised Verb Second effect;whether inversion will be triggered or not depends, in English as inGerman, on the question of whether the disjunction particle isnegative or not.

The facts in (70) thus present a context in which it can bedemonstrated that English is not the only Germanic language thatdistinguishes negative initial constituents from other initial material:Negative Inversion exists in German as well (though in German it isfar more difficult to diagnose because of the generalised Verb Secondfound in (non-coordinated) root clauses, which obliterates the effectof Negative Inversion in most contexts).31

Since whether, neither and nor (and nor’s German counterpartnoch in (70b) as well) cannot have been base-generated in their IP-external surface positions in (60) and (69b) (because that wouldprevent them from associating with the contrastive focus), theymust have been moved into the positions in the left periphery thatthey appear in. This, then, is a clear indication that whether, neitherand nor are phrasal, like either, the phrasality of which was de-monstrated in section 4.1.

4.2.4. Phrasal Movement of or: Locality RestrictionsIf nor is a phrasal category, as (69) demonstrates, then the sameshould arguably be true for or – unless or and nor are treated asprofoundly different lexical categories, an approach that certainly willnot qualify as the null hypothesis. That or is indeed phrasal andoriginates (like either ) close to the contrastive focus, rather than as adisjunction particle, is evidenced by two sets of empirical data

31 In older varieties of the Indo-European languages (e.g., Old and MiddleEnglish, Old Norse, Old, Middle and Early Modern High German, Middle and Early

Modern Dutch, and Old French; see Zwart 2005:section 3.2.2 and references citedthere), non-negative conjunction and disjunction particles could trigger inversion,giving rise to a generalised Verb Second pattern with the con/disjunction particles

raising to SpecCP.

MARCEL DEN DIKKEN

involving locality effects that manifest themselves in the context of the‘either too low’ puzzle.32

The first we encountered already at the end of section 3.5. Therewe saw that R-either across a sentential negation is impossible, asshown by the ungrammaticality of (47a), repeated below along withits grammatical counterparts in (47b,c).

(47)a. * John didn’t eat either RICE or he didn’t eat BEANS.

b. John either didn’t eat RICE or he didn’t eat BEANS.

c. Either John didn’t eat RICE or he didn’t eat BEANS.

We had also seen that the ban on R-either across a sentential negationcannot be blamed on either itself: either, after all, cannot floatrightward (i.e., downward); it is base-generated in the low position itis occupying in (47a), and in that low position, directly on the con-trastive focus, it would appear to be perfectly legitimate. The problemwith (47a), therefore, does not lie with either. Instead, I will arguethat it is or that is to blame.

If, as I argued extensively throughout the paper, either is aphrasal category and if nor is, too (cf. section 4.2.3), then weshould seriously entertain the possibility that or is a phrasal cat-egory as well – a phrasal category of the same type as either.Suppose, in particular, that or originates either directly on thesecond disjunct or alternatively within the second disjunct, at-tached to the contrastive focus or to a node on the h-path pro-jected by the contrastive focus (cf. (32)).33 Suppose, furthermore,

32 The fact that or could trigger inversion in earlier varieties of English as well asother Indo-European languages (cf. fn. 31) further supports this conclusion. JosephEmonds’ (p.c.) observation that the coordinating particle in Japanese occurs between

the first and second con/disjuncts may suggest that in Japanese, too, this particle isphrasal, not the lexicalisation of the head ‘J’ in (3) (which, given the general head-final character of Japanese, would be expected to surface exclusively after the secondcon/disjunct).

33 That or cannot originate outside the second disjunct altogether (for instance, inthe first disjunct or in the structure above the disjunction construction) follows from

the discussion in section 5, where it is argued that or needs to check features againstJ, the abstract functional head of the disjunction construction (cf. (3)): or mustoriginate in a position c-commanded by J so that an Agree relationship between J

and or can be established.


that or originates within the second disjunct (rather than directlyon the second disjunct) if and only if either occupies a positionwithin the first disjunct: or is preferably base-generated on thesecond disjunct; but parallelism between the two disjuncts ensuresthat if either is somewhere inside the first disjunct, or’s base po-sition in the second will match either’s in the first. Assuming thismuch, what we are confronted with in the R-either case in (47a) isan underlying representation in which or is base-generated insidethe second disjunct, directly on the contrastive focus (i.e., in aposition adjoined to the maximal projection of BEANS). But clearly,it has not stayed there in (47a) – it cannot stay there, in fact(something which I will address in section 5.3 below): it mustsurface at the left edge of the IP that forms the second disjunct.On the assumption that or originates in the same hierarchicalposition within the second disjunct as either does in the first, andgiven that that position is an object-adjoined position, or musthave moved to the position in which it surfaces, via a phrasalmovement operation, as depicted in (71). The problem, however, isthat or CANNOT so move: in so doing, it crosses a sententialnegation, which sets up an inner island (cf. Ross 1984).

(71) … ori [IP he didn’t eat [DP ti [DP BEANS]]]

This locality problem is averted in (47b), where, even though it onceagain has to be base-generated inside the second disjunct (becauseeither is embedded within the first), or originates higher up the treethan the sentential negation (in the same relative position that eitheroccupies). And naturally, no inner island effect arises in (47c) either:here, or attaches directly to the second disjunct and therefore doesnot need to move. But inner island problems are unavoidable in (47a)on present assumptions – and the fact that they do indeed manifestthemselves thus supports the analysis of or as a phrasal category thatmoves to the left edge of the second disjunct whenever it does notoriginate there.

A second context in which we can see phrasal movement of or atwork involves R-either cases in which either surfaces inside a PP orcomplex noun phrase. Consider first the case of PP-embedding.While (72a) reconfirms the fact (demonstrated already in section 3.1;cf. (21)) that either is in principle embeddable in a complement PP,

m ||

MARCEL DEN DIKKEN

the sentences in (72b,c), involving disjunction at the level of PP orIP, respectively, show that R-either inside a PP is impossible.34

(72)a. <Either> this sentence is <either> from <either>[NORWEGIAN] or [SWEDISH].

b. <Either> this sentence is <either> [from <*either>NORWEGIAN] or [from SWEDISH].

c. <Either> [this sentence is <either> from <*either>NORWEGIAN] or [it is from SWEDISH].

The examples in (73) make the same point: embedding either insidethe complement PP is legitimate if and only if disjunction is at thelevel of P’s complement (as in the a-examples); as soon as thedisjuncts are as large as PP or larger, either is prevented fromoccurring inside PP.

(73)a. <Either> John was <either> reading <either> from<either> [a BOOK] or [a MAGAZINE].

b. <Either> John was <either> reading <either> [from<*either> a BOOK] or [from a MAGAZINE].

c. <Either> John was <either> [reading <either> from<*either> a BOOK] or [reading from a MAGAZINE].

d. <Either> [John was <either> reading <either> from<*either> a BOOK] or [he was reading from a MAGAZINE].

The quintuplet in (74), which features the PP as the complement of anoun, adds one further piece to the puzzle. These sentences not onlyconfirm that R-either inside PP is impossible unless disjunction is atthe level of P’s complement, but they also show that embedding

34 The ungrammaticality of (72b,c) and (73b–d) with R-either in PP is reminiscentof Kayne’s (1998) discussion of sentences like John looked at only Mary all night,which are ungrammatical for many (though not all) speakers. Recall section 3.1. A

reviewer notes that some speakers even reject R-either to a P-less direct object (as inHe ate either RICE or he ate BEANS), which presumably relates to the fact that focusparticles to the immediate left of the direct object are sometimes deemed marked as

well (Bayer 1996:53).


R-either in a complex noun phrase fails as soon as the disjuncts arelarger than the noun phrase containing either.35

(74)a. <Either> John was <either> reading <either> a book<?either> about <either> [CHOMSKY] or [CHOPSTICKS].

b. <Either> John was <either> reading <either> a book<either> [about <*either> CHOMSKY] or [about CHOP-

STICKS].

c. <Either> John <either> was reading <either> [a book<?either> about <*either> CHOMSKY] or [a book aboutCHOPSTICKS].

d. <Either> John was <either> [reading <either> a book<*either> about <*either> CHOMSKY] or [reading abook about CHOPSTICKS].

e. <Either> [John was <either> reading <either> a book<*either> about <*either> CHOMSKY] or [he wasreading a book about CHOPSTICKS].

The account of the ‘either too low’ puzzle outlined in the precedingsections cannot accommodate the ungrammaticality of the relevantvariants of (72)–(74). Either does not seem to be at fault: it certainly isnot embedded within the focused constituent. But in either . . . or. . .disjunctions, it takes two to tango – when something goes awry, thatmay be because either trespassed, but as we have just discovered inour account of (47a), it may also be that or has made an illegitimatemove. The latter is indeed the case in (72)–(74), as I will argue in theremainder of this section.

I have argued that or is a phrasal category originating within thesecond disjunct iff its counterpart either originates in the first disjunct,with or’s base position in the second matching either’s in the first.Concretely, then, in (75a), since either is generated on the directobject of the first disjunct, or originates on the direct object of thesecond disjunct. From there, or moves to the position in which itsurfaces, via a phrasal movement operation, as depicted in (75b).

35 The variants of (74a) and (74c) with either placed on the about-PP are some-

what marginal. McCawley (1996:190) notes the same thing for only – cf. McCawley’sPupils only of HOROWITZ are allowed to play this piano and Books only about PHO-

NOLOGY don’t sell well, which he finds possible ‘[a]t least marginally’ (he does not give

them a question mark, though).

MARCEL DEN DIKKEN

(75)a. [IP John was reading either a BOOK] [or [IP he was reading aMAGAZINE]].

b. … ori [IP he was reading [DP ti [DP a MAGAZINE]]]

As in the case of (47a)/(71), we expect movement of or to be subject tothe rules of movement. And it is this expectation that then leads usback to the facts in (72)–(74). Let me pick out the most complicatedexample of the set, the one in (74e), and flesh out its structure in sucha way that it incorporates the insight that or originates in the samestructural position in the second disjunct that either originates in inthe first, and that, if not base-generated there, or ends up at the leftperiphery of the second disjunct via movement, which leaves behind atrace. The result is (74e¢ ).

(74e¢) <Either> [John was <either> reading <either> a book<*either> about <*either> CHOMSKY] ori [he was<Xti> reading <Xti> a book <*ti> about <*ti>CHOPSTICKS].

It will now be evident why the variants of (74e) featuring either to theright of book or about are ungrammatical: they are bad because, inthe derivation of these sentences, or must be extracted from a com-plex noun phrase, a book about CHOPSTICKS. Though wh-movementfrom a complex noun phrase of this type is possible in English, we arenot dealing with wh-movement in the case of the fronting of or.Instead, what we have on our hands is more like scrambling inlanguages like Dutch or German, which is known not to be able totransgress the boundaries of a complex noun phrase.36 So or cannotget out of the complex noun phrase into the clause-initial positionthat we see it in (74e). By the same token, (74d) also fails with either –and hence or ’s base position, too – to the right of book or about: hereagain, or originates inside the complex noun phrase headed by book,and ends up trapped right there.

m

36 Cf. Dutch (ia,b).

(i)a. dat Jan <*er> [het boek <er> over] gelezen heeft. (Dutch)

that Jan there the book there about read has

b. dat Jan <*erover> [het boek <erover>] gelezen heeft.

that Jan there-about the book there-about read has

that Jan read the book about it.


In (74c), where disjunction is at the level of the complex nounphrase, there is no extraction from a complex noun phrase going onin the derivation of any of the surface variants of the sentence. Witheither attached to the PP headed by about, or needs to move to the leftedge of the complex noun phrase – but in so doing, it does not extractfrom the noun phrase, and as a result, (74c) with either to the right ofbook is grammatical (albeit somewhat marked; recall also fn. 35).That the variant of (74c) with either (and hence or’s base position aswell) inside the about-PP is not well-formed is part and parcel of abroader generalisation that also covers the ungrammatical versions of(72b,c), (73b–d) and (74b) – a generalisation that is to be cast in termsof the way extraction from PP takes place.

In particular, the key to the analysis of the PP-containmentcases lies in the fact that extraction from PPs proceeds via anintermediate touch-down in an A¢-specifier position at the left edgeof the PP. Van Riemsdijk (1978) made the initial argument for thiskind of analysis on the basis of facts from Dutch and German. Inmore recent work on the syntactic structure of adpositionalphrases, Koopman (1997) and Den Dikken (2003) have arguedthat adpositions have a fully extended projection reaching all theway up to the CP-level (Koopman’s ‘CP(Place)’, for locative PPs).The specifier position of this CP is the escape hatch through whichcertain subconstituents of PP can make their way out of the PP –but the escape hatch position is not available for just anything. InDutch and German, the position is specifically restricted to the so-called [+R] pronouns (Van Riemsdijk 1978). In English, the set ofelements eligible for movement through PP’s escape hatch issomewhat larger but still not unrestricted: PPs, for instance, cannotextract from the larger PPs that contain them (cf. Who did you goaway after speaking to? vs. *To whom did you go away afterspeaking?). I will assume that the escape hatch position on the edgeof P’s extended projection is reserved, presumably universally, fornominal material (and in some languages, such as Dutch andGerman, for only a small subset thereof). Though establishing thecategorial status of or is not an entirely straightforward matter, itseems plain that it is not a nominal element: it resists articles ofany kind, as well as attributive modifiers, and will not sit in NPpositions by itself. Since, as a result of its non-nominality, or is notwelcome to the escape hatch at the left edge of the PP, it cannotmake its way to the left edge of the second disjunct in the variants

MARCEL DEN DIKKEN

of the examples in (72b,c), (73b–d) and (74b,c) in which either (andhence also or’s base position) is inside PP. Since or has to make itsway into that position (for reasons that I will address in section5.3), the relevant versions of these sentences crash, as desired.

This account of the facts in (72)–(74), just as that of theungrammaticality of (47a), crucially exploits movement – butmovement not of either but of or. We had already seen that nor, thenegative counterpart of or, must be able to undergo phrasal move-ment: it brings about Negative Inversion. We have now encounteredtwo concrete pieces of support for the view that or, like nor, is aphrasal constituent that undergoes leftward movement in the overtsyntax – movement which, whenever it is blocked by general localityrestrictions, causes the derivation to crash.

5. THE REPRESENTATION AND DERIVATION OF COORDINATE STRUCTURES

The discussion in section 4 emphasises the significance of thephrasal character of (n)either, whether and (n)or and of themovements that these elements undergo in the course of the syn-tactic derivation. But it still begs the question of why or shouldfront. This question will be at centre stage in section 5.3. But first,let me make it explicit what the foregoing discussion leads us to asfar as the syntactic structure underlying coordinate constructions isconcerned, and let me broaden the discussion to include conjunctionconstructions with both . . . and . . . as well. This is the topic ofsections 5.1 and 5.2.

5.1. The Syntactic Representation Underlying Coordinate Structures

The representation in (76a) (a lightly elaborated version of (3), above)illustrates the structure of (n)either . . . (n)or . . . coordination adoptedin this paper; (76b) is the structure of whether . . . or . . . construc-tions.37 These structures incorporate the idea that coordinate struc-tures are well-behaved binary-branching structures projected by a

37 Recall from the discussion in section 3 that either does not have to be base-generated in or directly on the first disjunct: in contexts in which a h-path can beprojected well beyond the first disjunct (as, for instance, in sentences such as those in

(1)), either can be base-generated anywhere along the h-path, which means that itmay also originate outside the disjunction. This is why the structure in (3)/(76a)includes a token of (n)either outside JP, a token which represents all cases in which

(n)either originates outside the disjunction structure, further up the h-path.


functional head (variously called ‘B’ (for ‘Boolean’; Munn 1993) or‘Co’ or ‘&’; I will label it ‘J’ (for ‘junction’) to cover both conjunctionand disjunction), but they do not assume, for English disjunctionconstructions, that the functional head that mediates between the twodisjuncts is lexically realised. That is to say, my analysis assumes thatin English ‘J’ is always empty in disjunction constructions: Englishlacks an overt disjunction; (n)or is not a lexicalisation of ‘J’ but in-stead a subconstituent of the second disjunct ( just like (n)either is asubconstituent of the first disjunct).

(76)a. <(n)either> (. . .) [JP [XP (. . .) <(n)either> . . .] [J [YP

(n)or . . .]]] (‘J’ = ‘(dis/con)junction’)

b. [CP whetheri (. . .) <ti> (. . .) [JP <ti> [XP . . .] [J [YP or. . .]]]]

c. <both> (. . .) [JP [XP (. . .) <both> . . .] [J [YP and . . .]]]

For conjunction constructions, an approach to both . . . and . . . thatassimilates it to (n)either . . . (n)or . . ., as in (76c), is highly plausibleas a null hypothesis. That and does indeed behave like (n)or in get-ting to the left edge of the second conjunct by movement if it does notoriginate there can be ascertained by looking at the facts discussed forR-either in section 4.2.4. In the interest of space, I confine myself tothe complex noun phrase examples in (74), which, in the domain ofboth . . . and . . ., present the same picture:

(77)a. John was reading both a book about Chomsky and he wasreading a book about chopsticks.

b. *John was reading a book both about Chomsky and he wasreading a book about chopsticks.

c. *John was reading a book about both Chomsky and hewas reading a book about chopsticks.

The grammaticality of (77a) shows, first of all, that R-both ispossible, like R-either. But more interestingly, the fact that (77b,c)are bad confirms that R-both is subject to the same restrictions asR-either. In section 4.2.4, I blamed the deviance of the either . . . or. . . cases corresponding to (77b,c) on the failure on the part of or tomake its way up to the left edge of the second disjunct in keepingwith the locality conditions on movement. If that analysis is on theright track, the fact that (77) mimicks the relevant variants of (74)

MARCEL DEN DIKKEN

then suggests that and is like or in being base-generated in a phrasalnode and undergoing leftward movement, to the edge of the secondconjunct.

5.2. L-either vs. L-both and the Status of Either and Both

Note, however, that both seems to be quite a bit less overtly phrasalthan (n)either in that both does not ‘run away from home’ as easily asdoes (n)either:

(78)a. John ate both rice and beans.

b. ?John both ate rice and beans.

c. *Both John ate rice and beans.

The sentence in (78b) is not quite as good as John either ate rice orbeans.38 Blodgett and Boland’s (1998) analysis of the first 150 tokensof both in the Wall Street Journal database of the PennTreebankCorpus did not bring forth a single case in which both has ‘run away’to the VP (though they did find one interesting example in which bothhas ‘run away’ to the DP: both the FTC and Justice Department); sothe phenomenon seems fairly marginal. Interestingly, however, intheir experimental work done on both in coordinate sentences,Blodgett and Boland (1998) and Blodgett (2000) found that L-bothcases of the type in (78b) are in fact quite popular in sentence-com-pletion tasks – thus, sentence onsets such asMary both sold vegetablesand . . . were completed by supplying a noun phrase following andapprox. 80% of the time. Of course this particular sentence com-pletion experiment may have biased L-both : inserting a noun phrase

38 Some ‘Googled’ examples of the type in (78b) are reproduced in (i).

(i)a. He both played the violin and Sinekeman [i.e., a Turkish musicalinstrument]. [http://interactive.m2.org/Music/cakin.html]

b. Nick knew that he both played football as well as basketball.

[http://www. angelfire.com/band2/backstreetfics/Downtime 11.html]

c. He both played as member and a leader.

[http://www.crosbystillsnash.com/csny/tour/05012002_PARTING_WORDS/05012002_reviewsPhotos.html]

d. He both studied at Yale Law School and for the Catholic priesthood

at the Sacred Heart Novitiate, a Jesuit seminary.[http://www.gracecathedral.org/enrichment/interviews/int_19991029i.shtml]


is obviously easier (i.e., less time- and energy-consuming) than sup-plying a VP. But in a follow-up judgement task run by Blodgett(2000), L-both was also judged acceptable to a significant degree(approx. 50% of the respondents deemed it perfectly fine), and in anexperiment testing listeners’ perception of prosodic cues a bias toidentify fragments as conjoined noun phrases emerged as well. Thephenomenon of L-both thus seems to be real, though it is presumablyfair to say that it is by no means as natural as L-either.

But whatever the status of L-both in (78b), it is clear that bothcategorically resists placement in sentence-initial position, as in (78c)(cf. also Larson 1985:260; Hendriks 2003:10). Notice, though, thatplacement of both in sentence-initial position is bad even in cases ofIP-coordination such as (79). The deviance of (78c) has nothing todo, therefore, with ‘migration’: both just does not want to be in thisposition. That both ’s aversion to this position does not have anythingto do with its being the initial position of a finite clause is evidentfrom the fact that it manifests itself also in (80b), where both imme-diately precedes a non-finite IP.

(79)a. Either John ate rice or Mary ate beans.

b. *Both John ate rice and Mary ate beans.

(80)a. The boys seem either [IP all t to have gone out] or [IP t tohave sneaked into the attic].

b. *The boys seem both [IP all t to have gone out] and [IP t tohave left their coats behind].

It is conceivable that this placement difference between either andboth is related to the fact that both qua quantifier of noun phrasesdemands a plural host (cf. (81c)).

(81)a. [Either analysis] will yield the desired result.

b. [Neither analysis] will yield the desired result.

c. [Both analyses] will yield the desired result.

The fact that IP has no interpretable number feature may be a factorcontributing to the ungrammaticality of (78c), (79b) and (80b)(though Aniko Liptak, p.c., is right to point out that the grammati-cality of This toy is both expensive and useless then stands out as amystery: APs presumably have no interpretable number features

MARCEL DEN DIKKEN

either). If indeed there is a way of making sense of (78c), (79b) and(80b) along these lines, it probably lends support to the hypothesisthat both and (n)either in con/disjunctions are not con/disjunctionparticles but quantifiers (see also Dougherty 1970; Higginbotham1991; Munn 1993; Culicover 1999 for the observation that thedisjunctions either and neither double as quantifiers).39

5.3. Why (n)or and and must be Initial in the Second Disjunct

This said, let me return to the question of what forces the (n)or/andin the second dis/conjunct to front to the left edge of the dis/conjunct.

39 Aniko Liptak (p.c.) points out that the Hungarian counterpart to both . . . and. . . (mind . . . mind . . .) employs a quantificational element: mind is the floating uni-

versal quantifier ‘all’. This supports the text analysis. But interestingly, even thoughHungarian mind . . . mind . . . admits of exactly two conjuncts, it does not actually use(floating) ‘both’ (i.e., mindket). Note that in English as well, there are distributional

parallels between L-either and floating quantifiers: thus, the answer in (iA) isungrammatical if cleverly is interpreted as a manner adverb (‘in a clever way’) butfine with cleverly read as a subject-oriented adverb (‘it was clever of her to. . .’ or ‘shewas clever enough to. . .’); the fact that L-either resists being placed below a manner

adverb matches the fact that floating all cannot be placed below a manner adverbeither, as seen in (ii) (Bobaljik 1995), which suggests that attachment of floatingquantifiers and L-either to lexical projections is impossible (but see Boskovic 2004 for

a different interpretation of (ii)).

(i)Q: John discussed the English and French facts. What did Sue do?

A: She cleverly <%either> discussed the DUTCH facts or the GERMAN

facts.

(ii) The students <all> completely <*all> understood.

That (n)either/both qua quantifier differs from (n)either/both qua dis/conjunctionin being unable to ‘run away from home’ (cf. John<*either/*both > likes <either/both> reindeer) does not necessarily defeat an assimilation of the two elements: thefact that quantificational (n)either/both cannot move away from its nominal host

reduces to the entirely general fact that, in languages like English, left-branch con-stituents of noun phrases cannot be extracted (cf. the Left Branch Condition).Potentially more challenging differences between either qua ‘disjunction’ and either

qua quantifier are the fact that the former is not necessarily binary (cf. John ate eitherrice or beans or potatoes; contra Sag et al. 1985) and that, while the former has a[+WH] counterpart whether, there is no such thing as [QP whether analysis ] serving

as a wh-phrase in Modern English (though there was in pre-17th c. English, and inbiblical English through the 19th c.; see Jespersen 1961:Vol. II, p. 200: Whether ofthem twaine did the will of his father?); instead of whether, Modern English uses which

in this context. I have no solutions to offer for these conundrums.


We have seen that (n)either and whether can be base-generated in avariety of positions on the h-path, and that, from there, neither hasthe option and whether the obligation to move to higher positions, inkeeping with locality. But (n)or and and have no choice but to be atthe left edge of the second dis/conjunct – that is, examples of the typein (82) and (83) are systematically impossible.

(82)a. *[[John <either> ate <either> rice] [JB [he <or>ate<or>beans]]]

b. *[[John <neither>ate <neither> rice] [JB [he <nor> ate<or> beans]]]

c. *[[John <both> ate <both> rice] [JB [he <and> ate<and> beans]]]

(83)a. *[[John either laughed] [JB [he or cried]]]

b. *[[John neither laughed] [JB [he nor cried]]]

c. *[[John both laughed] [JB [he and cried]]]

Regardless of the locus of focus, (n)or/and must appear at the leftedge of the second dis/conjunct. Two possible hypotheses spring tomind when it comes to answering the question of why this should bethe case:

(i) (n)or/and has to raise up to JB in overt syntax to licenseit(s emptiness); OR:

(ii) (n)or/and has to raise to the edge of the second disjunct (a‘phase’ in the sense of Chomsky 2001) so that JB canestablish an Agree relationship with (n)or/and, checking itsfeatures.

Of these two hypotheses, the latter is presumably the right one forthe cases at hand – particularly in view of the fact that what weknow about the vicissitudes of head movement (cf. Baker 1988;Hale and Keyser 1993 etc.) suggests that it should be very difficultfor (n)or/and, which after all is a phrasal constituent that occupiesa left-branch adjunction position in the tree, to make its way up toJ via head movement. I therefore opt for the scenario in (ii) andassume that J has a bundle of uninterpretable formal featureswhich can only be checked against a matching bundle of formalfeatures in (n)or/and – which explains why (n)or/and must occur

MARCEL DEN DIKKEN

on the second dis/conjunct, and cannot occur elsewhere (recall fn.33). On the further assumption that any complement of J that hasan instance of (n)or/and in it is systematically a phase, the require-ment that J establish an Agree relationship with an embedded (n)or/andwill then force (n)or/and to raise to the edge of the phase if it is notbase-generated there – which will make it end up in initial position, asdesired.

Thus, base-generating (n)or/and as a phrasal constituent has noadverse consequences for the account of the fact that (n)or/andmust be initial in the second dis/conjunct. And of course the ideathat (n)or and and originate in a phrasal position has the significantadvantages of (a) assimilating them to their correlates (n)either andboth in/on/outside the first dis/conjunct, (b) explaining the fact thatfronting of negative nor necessarily leads to subject-auxiliaryinversion (an instance of Negative Inversion), and (c) accounting forthe locality and intervention effects canvassed in section 4.

5.4. Acquisition

In closing, let me raise the question of how the learner shouldreach the conclusion that the junction head ‘J’ in English coordi-nation constructions is systematically empty and that the overtcoordinator (or, nor, and) originates inside the second dis/conjunct.For nor, the answer to this question is most easily reached, on thebasis of the positive evidence from Negative Inversion in the sec-ond disjunct. The fact that nor must be initial while neither can butdoes not have to be then leads to the conclusion that there is somestructural trigger for placing the coordinator at the left edge of thesecond disjunct (the trigger being the need to establish a feature-checking Agree relationship with ‘J’), in the same position thatneither moves into when it does indeed front. Extrapolating fromthe fact that nor can end up at the left edge of the second disjunctvia movement and must hence be phrasal, the learner reaches thesame conclusion (in the absence of evidence to the contrary) for orand, beyond this, and as well. This delivers precisely the result wedesire – a grammar that incorporates (3)/(76) that successfullygenerates the well-formed cases of L-either/both and R-either/boththat were catalogued in this paper and rules out all the ungram-matical cases (which, given the ‘no negative evidence’ constraint,cannot be the learner’s impetus to the construction of the grammarof dis/conjunction).


6. CONCLUDING REMARKS

Starting out from the two major puzzles that the syntax of either . . .or . . . presents us with – the ‘either too high’ puzzle in (1) and the‘either too low’ puzzle in (2) – this paper has laid out an extendedargument in favour of an approach to either . . . or . . . that assumesthat (i) both either and or are phrasal categories, (ii) either and or arebase-generated in a position adjoined directly to their disjunct or to (anode on the h-path projected from) the contrastive focus, but neverlower than (or inside) the contrastive focus, and (iii) (n)or, if notbase-generated on the edge of the second disjunct, moves there(subject to the familiar restrictions on movement) in order to be ableto participate in a feature-checking Agree relationship with theabstract head J(unction), the head that takes the second disjunct as itscomplement and the first disjunct as its specifier, as depicted in (3).The analysis was extended to cover whether . . . or . . . and both . . . and. . . coordination constructions, which likewise feature the JP struc-ture in (3) (cf. (76)) and again do not analyse the element introducingthe second dis/conjunct as the lexicalisation of the ‘J’ head but as aphrasal category establishing a feature-checking relationship withabstract ‘J’ instead.

REFERENCES

Abney, Stephen. 1987. The English Noun Phrase in its Sentential Aspect, Ph.D.dissertation, MIT.

Baker, Mark. 1988. Incorporation, University of Chicago Press, Chicago.

Bayer, Josef, 1996. Directionality and Logical Form, Kluwer Academic Publishers,Dordrecht.

Blodgett, Allison. 2000. ‘Prosodic Cues ‘‘and’’ Syntactic Disambiguation’, OSU

Working Papers in Linguistics 53, 1–30.Blodgett, Allison and Julie Boland. 1998, ‘Understanding both in CoordinateStructures: The OSU Subject Pool Both Surprised Allison and Julie!’, Poster

presented at the 11th CUNY Conference on Human Sentence Processing, RutgersUniversity, March 1998.

Bobaljik, Jonathan. 1995. Morphosyntax: The Syntax of Verbal Inflection, Ph.D.dissertation, MIT.

Boskovic, Zeljko, 2004. ‘Be Careful Where you Float your Quantifiers’, NaturalLanguage and Linguistic Theory 22, 681–742.

Buring, Daniel. 2002. ‘Focus Projection and Default Phonology’, Paper presented at

Lund University, December 2002; ms., UCLA.Chomsky, Noam. 1971. ‘Deep Structure, Surface Structure and Semantic Interpre-tation’, in D. Steinberg and J. Jakobovits (eds.), Semantics: An Interdisciplinary

MARCEL DEN DIKKEN

Reader in Philosophy, Linguistics and Psychology, Cambridge University Press,Cambridge, pp. 183–216.

Chomsky, Noam. 1986. Barriers, MIT Press, Cambridge, MA.Chomsky, Noam. 1995. The Minimalist Program, MIT Press, Cambridge, MA.

Chomsky, Noam. 2001. ‘Derivation by Phase’, in M. Kenstowicz (ed.), Ken Hale : ALife in Language. MIT Press, Cambridge, MA, pp. 1–52.

Culicover, Peter. 1999. Syntactic Nuts, Cambridge University Press, Cambridge.

Den Dikken, Marcel. 1995. Particles. On the Syntax of Verb-Particle, Triadic andCausative Constructions, Oxford University Press, Oxford/New York.

Den Dikken, Marcel. 2003. ‘On the Syntax of Locative and Directional Adpositional

Phrases’, ms., CUNY Graduate Center.Dougherty, Ray. 1970. ‘A Grammar of Coordinate Conjoined Structures: I’,Language 46, 850–898.

E. Kiss, Katalin. 1998. ‘Identificational Focus Versus Information Focus’, Language74, 245–273.

E. Kiss, Katalin. 2000. The Syntax of Hungarian, Cambridge University Press,Cambridge.

Emonds, Joseph. 1985. A Unified Theory of Syntactic Categories, Foris, Dordrecht.Gullı, Antonino. 2003. Phrasal Reduplication in Syntax, Ph.D. dissertation, CUNYGraduate Center.

Hale, Kenneth and Samuel Jay Keyser. 1993. ‘On Argument Structure and theLexical Expression of Syntactic Relations’, in K. Hale and S.J. Keyser (eds.), TheView from Building 20, MIT Press, Cambridge, MA, pp. 55–109.

Han, Chung-hye and Maribel Romero. 2004. ‘The Syntax of Whether/Q . . . orQuestions’, Natural Language and Linguistic Theory 22 , 527–564.

Hendriks, Petra. 2001. ‘Initial Coordination and the Law of Coordination of Likes’,in H. Broekhuis and T. van der Wouden (eds.), Linguistics in the Netherlands 2001,

John Benjamins, Amsterdam, pp.127–138.Hendriks, Petra. 2003. ‘Either’ as a Focus Particle’, ms., University of Groningen.Hendriks, Petra. 2004. ‘Either, Both and Neither in Coordinate Structures’, ms.,

University of Groningen.Higginbotham, James. 1985. ‘On Semantics’, Linguistic Inquiry 16, 547–593.Higginbotham, James. 1991. ‘Either/or’, in T. Sherer (ed.), Proceedings of the North

East Linguistic Society, Vol. 21, GLSA Publications, Amherst, MA, pp. 143–155.Hoeksema, Jack and Frans Zwarts. 1991. ‘Some Remarks on Focus Adverbs’,Journal of Semantics 8, 51–70.

Hoekstra, Teun. 1988. �Small Clause Results’, Lingua 74 , 101–139.Jackendoff, Ray. 1972. Semantic Interpretation in Generative Grammar, MIT Press,Cambridge, MA.

Jespersen, Otto. 1961. A Modern English Grammar on Historical Principles, George

Allen & Unwin London. Ejnar Munksgaar, Copenhagen.Johannessen, Janne Bondi. 1998. Coordination, Oxford University Press, Oxford/New York.

Johannessen, Janne Bondi. 2005. ‘The Syntax of Correlative adverbs’, Lingua 115,419–443.

Kayne, Richard. 1975. French Syntax. The Transformational Cycle, MIT Press,

Cambridge, MA.Kayne, Richard. 1984. Connectedness and Binary Branching, Foris, Dordrecht.


Kayne, Richard. 1994. The Antisymmetry of Syntax, MIT Press, Cambridge, MA.Kayne, Richard. 1998. ‘Overt vs. Covert Movement’, Syntax 1 , 128–191. (Reprintedin Richard Kayne, 2000, Parameters and Universals, Oxford University Press,Oxford/New York, pp. 223–281.)

Kenesei, Istvan.1986. ‘On the Logic of Hungarian Word Order’, in W. Abraham andS. de Meij (eds.), Topic, Focus, and Configurationality, John Benjamins, Amster-dam, pp. 143–159.

Klima, Edward. 1964. ‘Negation in English’, in J. Fodor and J. Katz (eds.), TheStructure of Language: Readings in the Philosophy of Language, Prentice-Hall,Englewood Cliffs, NJ, pp. 246–323.

Koopman, Hilda. 1997. ‘Prepositions, Postpositions, Circumpositions and Particles:The Structure of Dutch PPs’, ms., UCLA (available on-line at: http://www.lin-guistics.ucla.edu/people/koopman/papers/pp.pdf).

Larson, Richard. 1985. ‘On the Syntax of Disjunction Scope’, Natural Language andLinguistic Theory 3, 217–264.

Lechner, Winfried. 2000. ‘Bivalent Coordination in German’, Snippets 1, 11–12.Liptak, Aniko. 2001. On the Syntax of Wh-items in Hungarian, Ph.D. dissertation,

University of Leiden.Manzini, Rita and Anna Roussou. 2000. ‘A Minimalist Theory of A-movement andControl’, Lingua 110, 409–447.

Marantz, Alec. 1997. ‘No Escape from Syntax: Don’t Try Morphological Analysis inthe Privacy of Your Own Lexicon’, in A. Dimitriadis et al. (eds.), Penn WorkingPapers in Linguistics, Vol. 4, pp. 201–225.

McCawley, James. 1996. ‘The Focus and Scope of Only’, in B. Partee and P. Sgall(eds.), Discourse and Meaning. Papers in Honor of Eva Hajicova, John Benjamins,Amsterdam, pp. 171–193.

Munn, Alan. 1993. Topics in the Syntax and Semantics of Coordinate Structures,

Ph.D. dissertation, University of Maryland, College Park.Neijt, Anneke. 1979. Gapping: A Contribution to Sentence Grammar, Foris,Dordrecht.

Radford, Andrew. 1997. Syntactic Theory and the Structure of English: A MinimalistApproach, Cambridge University Press, Cambridge.

Rochemont, Michael. 1986. Focus in Generative Grammar, John Benjamins,

Amsterdam.Rooth, Mats. 1985. Association with Focus, Ph.D. dissertation, University of Mas-sachusetts, Amherst.

Ross, John Robert. 1984. ‘Inner Islands’, in C. Brugman and M. Macaulay (eds.),Proceedings of the 10th Annual Meeting of the Berkeley Linguistics Society, pp.258–265.

Sag, Ivan, Gerald Gazdar, Thomas Wasow and Steven Weisler. 1985. ‘Coordination

and How to Distinguish Categories’, Natural Language and Linguistic Theory 3,117–171.

Schwarz, Bernhard. 1999. ‘On the Syntax of Either . . . or’, Natural Language and

Linguistic Theory 17, 339–370.Schwarzschild, Roger. 1999. ‘GIVENness, AVOIDF and other Constraints on thePlacement of Accents’, Natural Language Semantics 7, 141–177.

Selkirk, Lisa. 1984. Phonology and Syntax: The Relation between Sound and Struc-ture, MIT Press, Cambridge, MA.

MARCEL DEN DIKKEN

Selkirk, Lisa. 1995. ‘Sentence Prosody: Intonation, Stress, and Phrasing’, inJ. Goldsmith (ed.), The Handbook of Phonological Theory, Blackwell, Oxford,pp. 550–569.

Selkirk, Lisa. 2003. ‘Focus Prosody: Interface Economy and Surface Enhancement’,

abstract of a paper presented at Yale University, available on-line at: http://www.yale.edu/linguist/SelkirkAbstract.doc.

Stowell, Tim. 1981. Origins of Phrase Structure, Ph.D. dissertation, MIT.

Van Riemsdijk, Henk 1978. A Case Study in Syntactic Markedness, The Peter deRidder Press, Lisse.

Wesche, Birgit, 1995. Symmetric Coordination. An Alternative Theory of Phrase

Structure. Linguistische Arbeiten, Max Niemeyer Verlag, Tubingen.Wilder, Chris. 1997. ‘Some Properties of Ellipsis in Coordination’, in A. Alexiadouand T.A. Hall (eds.), Studies on Universal Grammar and Typological Variation,

John Benjamins, Amsterdam, pp. 59– 107.Williams, Edwin. 1981. ‘Argument structure and morphology’, The Linguistic Review1 , 81–114.

Zwart, Jan-Wouter. 2005. ‘Verb Second as a Function of Merge’, in M. den Dikken

and C. Tortora (eds.), The Function of Function Words and Functional Categories,John Benjamins, Amsterdam, pp. 11–40.

Received 7 November 2003Revised 26 April 2005

Linguistics Program

CUNY Graduate Center365 Fifth AvenueNew York, NY 10016–4309USA

<[email protected]>


either-float and the syntax of co-€¦ · ellipsis approach to displacement of either (cf....

Documents